Gemma 4 31B

import { createClient } from '@runware/sdk'

const client = await createClient({ apiKey: process.env.RUNWARE_API_KEY })
await client.connect()

const [result] = await client.run({
  model: 'google:gemma@4-31b',
  inputs: {
    images: [
      'https://assets.runware.ai/assets/inputs/4827db34-f4f0-4277-a5fe-a2b2f7288672.jpg'
    ],
    videos: [
      'https://assets.runware.ai/assets/inputs/f62fa264-2764-4ec4-b1a4-70c366fec9fd.mp4'
    ]
  },
  messages: [
    {
      role: 'user',
      content: 'Act as a senior EHS investigator for a distribution center. Analyze the attached ground-level inspection photograph and CCTV video together, then draft a concise near-miss investigation report that management can use at tomorrow\'s safety review.\n\nUse only details visible in the supplied media. Clearly separate direct observations from reasonable inferences, and state when something cannot be determined. Review the full video sequence and cite approximate timestamps for key moments.\n\nStructure the report as follows:\n1. Executive summary: 100–150 words explaining what occurred, the potential consequence, and the most urgent control.\n2. Evidence log: a table identifying the source, timestamp or image region, objective observation, and safety relevance.\n3. Event sequence: a chronological reconstruction from approach through recovery.\n4. Contributing factors: organize under people, equipment, environment, procedure, and supervision; do not assign blame or claim root causes unsupported by the media.\n5. Risk assessment: rate potential severity and recurrence likelihood as Low, Medium, High, or Critical, with a brief rationale.\n6. Corrective action plan: a prioritized table with action, control hierarchy level, owner role, target timeframe, and verification method. Include immediate 24-hour controls, 30-day improvements, and longer-term engineering options.\n7. Five questions investigators should ask witnesses or supervisors to close evidence gaps.\n\nCross-reference the photograph with the video—for example, determine whether sight-line obstructions, floor markings, mirror placement, pallet storage, lighting, traffic separation, or behavior visible in one source helps explain events in the other. Do not identify individuals or infer protected personal characteristics.'
    }
  ]
})

import asyncio
import os

from runware import Runware


async def main():
    async with Runware(api_key=os.environ["RUNWARE_API_KEY"]) as client:
        results = await client.run({
            "model": "google:gemma@4-31b",
            "inputs": {
                "images": [
                    "https://assets.runware.ai/assets/inputs/4827db34-f4f0-4277-a5fe-a2b2f7288672.jpg"
                ],
                "videos": [
                    "https://assets.runware.ai/assets/inputs/f62fa264-2764-4ec4-b1a4-70c366fec9fd.mp4"
                ]
            },
            "messages": [
                {
                    "role": "user",
                    "content": "Act as a senior EHS investigator for a distribution center. Analyze the attached ground-level inspection photograph and CCTV video together, then draft a concise near-miss investigation report that management can use at tomorrow's safety review.\n\nUse only details visible in the supplied media. Clearly separate direct observations from reasonable inferences, and state when something cannot be determined. Review the full video sequence and cite approximate timestamps for key moments.\n\nStructure the report as follows:\n1. Executive summary: 100–150 words explaining what occurred, the potential consequence, and the most urgent control.\n2. Evidence log: a table identifying the source, timestamp or image region, objective observation, and safety relevance.\n3. Event sequence: a chronological reconstruction from approach through recovery.\n4. Contributing factors: organize under people, equipment, environment, procedure, and supervision; do not assign blame or claim root causes unsupported by the media.\n5. Risk assessment: rate potential severity and recurrence likelihood as Low, Medium, High, or Critical, with a brief rationale.\n6. Corrective action plan: a prioritized table with action, control hierarchy level, owner role, target timeframe, and verification method. Include immediate 24-hour controls, 30-day improvements, and longer-term engineering options.\n7. Five questions investigators should ask witnesses or supervisors to close evidence gaps.\n\nCross-reference the photograph with the video—for example, determine whether sight-line obstructions, floor markings, mirror placement, pallet storage, lighting, traffic separation, or behavior visible in one source helps explain events in the other. Do not identify individuals or infer protected personal characteristics."
                }
            ]
        })


asyncio.run(main())

curl https://api.runware.ai/v1 \
  -H "Authorization: Bearer $RUNWARE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '[
    {
      "taskType": "textInference",
      "taskUUID": "75d463c3-5515-428c-b85e-fa75e047f84b",
      "model": "google:gemma@4-31b",
      "inputs": {
        "images": [
          "https://assets.runware.ai/assets/inputs/4827db34-f4f0-4277-a5fe-a2b2f7288672.jpg"
        ],
        "videos": [
          "https://assets.runware.ai/assets/inputs/f62fa264-2764-4ec4-b1a4-70c366fec9fd.mp4"
        ]
      },
      "messages": [
        {
          "role": "user",
          "content": "Act as a senior EHS investigator for a distribution center. Analyze the attached ground-level inspection photograph and CCTV video together, then draft a concise near-miss investigation report that management can use at tomorrow's safety review.\n\nUse only details visible in the supplied media. Clearly separate direct observations from reasonable inferences, and state when something cannot be determined. Review the full video sequence and cite approximate timestamps for key moments.\n\nStructure the report as follows:\n1. Executive summary: 100–150 words explaining what occurred, the potential consequence, and the most urgent control.\n2. Evidence log: a table identifying the source, timestamp or image region, objective observation, and safety relevance.\n3. Event sequence: a chronological reconstruction from approach through recovery.\n4. Contributing factors: organize under people, equipment, environment, procedure, and supervision; do not assign blame or claim root causes unsupported by the media.\n5. Risk assessment: rate potential severity and recurrence likelihood as Low, Medium, High, or Critical, with a brief rationale.\n6. Corrective action plan: a prioritized table with action, control hierarchy level, owner role, target timeframe, and verification method. Include immediate 24-hour controls, 30-day improvements, and longer-term engineering options.\n7. Five questions investigators should ask witnesses or supervisors to close evidence gaps.\n\nCross-reference the photograph with the video—for example, determine whether sight-line obstructions, floor markings, mirror placement, pallet storage, lighting, traffic separation, or behavior visible in one source helps explain events in the other. Do not identify individuals or infer protected personal characteristics."
        }
      ]
    }
  ]'

runware run google:gemma@4-31b \
  inputs.images.0=https://assets.runware.ai/assets/inputs/4827db34-f4f0-4277-a5fe-a2b2f7288672.jpg \
  inputs.videos.0=https://assets.runware.ai/assets/inputs/f62fa264-2764-4ec4-b1a4-70c366fec9fd.mp4 \
  messages.0.role=user \
  messages.0.content="Act as a senior EHS investigator for a distribution center. Analyze the attached ground-level inspection photograph and CCTV video together, then draft a concise near-miss investigation report that management can use at tomorrow's safety review.

Use only details visible in the supplied media. Clearly separate direct observations from reasonable inferences, and state when something cannot be determined. Review the full video sequence and cite approximate timestamps for key moments.

Structure the report as follows:
1. Executive summary: 100–150 words explaining what occurred, the potential consequence, and the most urgent control.
2. Evidence log: a table identifying the source, timestamp or image region, objective observation, and safety relevance.
3. Event sequence: a chronological reconstruction from approach through recovery.
4. Contributing factors: organize under people, equipment, environment, procedure, and supervision; do not assign blame or claim root causes unsupported by the media.
5. Risk assessment: rate potential severity and recurrence likelihood as Low, Medium, High, or Critical, with a brief rationale.
6. Corrective action plan: a prioritized table with action, control hierarchy level, owner role, target timeframe, and verification method. Include immediate 24-hour controls, 30-day improvements, and longer-term engineering options.
7. Five questions investigators should ask witnesses or supervisors to close evidence gaps.

Cross-reference the photograph with the video—for example, determine whether sight-line obstructions, floor markings, mirror placement, pallet storage, lighting, traffic separation, or behavior visible in one source helps explain events in the other. Do not identify individuals or infer protected personal characteristics."

{
  "taskType": "textInference",
  "taskUUID": "75d463c3-5515-428c-b85e-fa75e047f84b",
  "model": "google:gemma@4-31b",
  "inputs": {
    "images": [
      "https://assets.runware.ai/assets/inputs/4827db34-f4f0-4277-a5fe-a2b2f7288672.jpg"
    ],
    "videos": [
      "https://assets.runware.ai/assets/inputs/f62fa264-2764-4ec4-b1a4-70c366fec9fd.mp4"
    ]
  },
  "messages": [
    {
      "role": "user",
      "content": "Act as a senior EHS investigator for a distribution center. Analyze the attached ground-level inspection photograph and CCTV video together, then draft a concise near-miss investigation report that management can use at tomorrow's safety review.\n\nUse only details visible in the supplied media. Clearly separate direct observations from reasonable inferences, and state when something cannot be determined. Review the full video sequence and cite approximate timestamps for key moments.\n\nStructure the report as follows:\n1. Executive summary: 100–150 words explaining what occurred, the potential consequence, and the most urgent control.\n2. Evidence log: a table identifying the source, timestamp or image region, objective observation, and safety relevance.\n3. Event sequence: a chronological reconstruction from approach through recovery.\n4. Contributing factors: organize under people, equipment, environment, procedure, and supervision; do not assign blame or claim root causes unsupported by the media.\n5. Risk assessment: rate potential severity and recurrence likelihood as Low, Medium, High, or Critical, with a brief rationale.\n6. Corrective action plan: a prioritized table with action, control hierarchy level, owner role, target timeframe, and verification method. Include immediate 24-hour controls, 30-day improvements, and longer-term engineering options.\n7. Five questions investigators should ask witnesses or supervisors to close evidence gaps.\n\nCross-reference the photograph with the video—for example, determine whether sight-line obstructions, floor markings, mirror placement, pallet storage, lighting, traffic separation, or behavior visible in one source helps explain events in the other. Do not identify individuals or infer protected personal characteristics."
    }
  ]
}

import { createClient } from '@runware/sdk'

const client = await createClient({ apiKey: process.env.RUNWARE_API_KEY })
await client.connect()

const [result] = await client.run({
  model: 'google:gemma@4-31b',
  messages: [
    {
      role: 'user',
      content: 'Act as the lead interpretive writer for a maritime museum. Create the final gallery text package for “North by Necessity,” a small exhibition about the 1884 rescue of the Greely Arctic expedition. The writing must be historically responsible, engaging for general visitors, and ready for curatorial review.\n\nUse only the verified dossier below. Do not introduce quotations, motives, dates, Indigenous names, technical specifications, or other facts that are not supplied. Where the evidence does not support certainty, use appropriately qualified language.\n\nVERIFIED DOSSIER\n- In 1881, U.S. Army First Lieutenant Adolphus Greely led 25 men to Lady Franklin Bay on Ellesmere Island as part of the First International Polar Year.\n- The party established Fort Conger and conducted meteorological, astronomical, magnetic, tidal, and geographic observations.\n- In 1882 and 1883, scheduled relief ships failed to reach the party because of ice and poor decisions within the relief efforts.\n- Following written orders, Greely’s party abandoned Fort Conger in August 1883 and traveled south. They eventually established a camp at Cape Sabine.\n- Food caches expected near Cape Sabine were inadequate. The party faced extreme hunger, cold, illness, isolation, and internal conflict.\n- Several earlier rescue attempts failed. In 1884, the U.S. Navy organized another expedition under Commander Winfield Scott Schley.\n- Schley’s squadron included the steam-powered ships Thetis, Bear, and Alert. Bear and Thetis reached Cape Sabine on June 22, 1884.\n- Rescuers found seven men alive; one of those seven, Joseph Elison, died during the return voyage. Of the original 25, six ultimately returned alive.\n- Recovery of scientific records was considered an important outcome of the expedition.\n- The expedition took place in an Arctic homeland that was already known, traveled, and inhabited by Indigenous peoples. Nineteenth-century U.S. accounts often described such regions as “unknown” or “unexplored,” language that reflected outsiders’ perspectives rather than an absence of existing knowledge.\n- The exhibition contains these four objects:\n  1. A brass-and-glass aneroid barometer used for field observation, maker unknown, circa 1880.\n  2. A water-stained page from a meteorological observation ledger kept at Fort Conger in February 1882; it records temperature, pressure, cloud cover, and wind.\n  3. A blue wool uniform coat attributed to Winfield Scott Schley and worn during the 1884 rescue expedition; attribution is supported by family provenance but is not conclusive.\n  4. An 1885 newspaper engraving depicting the survivors’ arrival in Portsmouth, New Hampshire. It was produced after the event for a mass audience and should be treated as a mediated representation, not an eyewitness record.\n\nAUDIENCE AND VOICE\n- Primary audience: visitors aged 14 and older with no prior knowledge of the event.\n- Reading level: approximately U.S. grade 8.\n- Tone: restrained, humane, vivid, and precise—not triumphalist, sensational, or melodramatic.\n- Center survival and scientific work without turning suffering into spectacle.\n- Do not frame the Arctic as empty, conquered, or waiting to be discovered.\n- Explain unfamiliar terms in context. Prefer short sentences and active voice.\n- Avoid rhetorical questions, exclamation points, invented scene-setting, and clichés such as “frozen wasteland,” “race against time,” or “against all odds.”\n- Refer to the group as “the Greely party,” not “Greely’s men,” except where grammar makes that impossible.\n\nDELIVERABLES\n1. Exhibition title and subtitle. Retain “North by Necessity” as the title and write a subtitle of no more than 12 words.\n2. Introductory panel: 170–190 words. Establish the scientific mission, failed relief efforts, move to Cape Sabine, rescue, human cost, and the need to question period ideas of “exploration.”\n3. Section panel, “Measuring the Arctic”: 105–125 words. Explain why sustained observations mattered and how routine recordkeeping continued under difficult conditions.\n4. Section panel, “Relief Delayed”: 105–125 words. Clarify that the disaster resulted from both environmental hazards and human decisions. Avoid assigning blame beyond the dossier.\n5. Section panel, “Rescue and Return”: 105–125 words. Describe Schley’s expedition, the June 22 rescue, the number of survivors, and the recovery of records without presenting the outcome as an uncomplicated victory.\n6. Four object labels, one for each listed object, each 65–80 words. Begin each label with a display heading of no more than six words. For the coat, clearly communicate the uncertain attribution. For the engraving, teach visitors why a published image is not neutral evidence.\n7. A family conversation prompt of 35–45 words that invites observation and discussion without using a rhetorical question. It should work beside the ledger and barometer.\n8. A 90–110 word accessibility-oriented visual description of the newspaper engraving. Because no image is provided, limit the description to what can responsibly be inferred from the dossier: medium, subject, publication context, and evidentiary limits. Explicitly state that composition and depicted details must be confirmed against the physical object before publication.\n9. A curator review table with columns: “Copy location,” “Claim or wording needing review,” and “Reason.” Flag at least four issues, including Indigenous context, the coat attribution, the engraving’s evidentiary status, and survivor totals.\n\nFORMAT\nPresent the finished package in the deliverable order above with clear headings. Put word counts in parentheses after every panel, label, prompt, and visual description. Do not include citations in visitor-facing copy. After the visitor-facing materials, include the curator review table and then a brief style-compliance checklist confirming that all required facts, cautions, and word limits have been addressed.'
    }
  ]
})

import asyncio
import os

from runware import Runware


async def main():
    async with Runware(api_key=os.environ["RUNWARE_API_KEY"]) as client:
        results = await client.run({
            "model": "google:gemma@4-31b",
            "messages": [
                {
                    "role": "user",
                    "content": "Act as the lead interpretive writer for a maritime museum. Create the final gallery text package for “North by Necessity,” a small exhibition about the 1884 rescue of the Greely Arctic expedition. The writing must be historically responsible, engaging for general visitors, and ready for curatorial review.\n\nUse only the verified dossier below. Do not introduce quotations, motives, dates, Indigenous names, technical specifications, or other facts that are not supplied. Where the evidence does not support certainty, use appropriately qualified language.\n\nVERIFIED DOSSIER\n- In 1881, U.S. Army First Lieutenant Adolphus Greely led 25 men to Lady Franklin Bay on Ellesmere Island as part of the First International Polar Year.\n- The party established Fort Conger and conducted meteorological, astronomical, magnetic, tidal, and geographic observations.\n- In 1882 and 1883, scheduled relief ships failed to reach the party because of ice and poor decisions within the relief efforts.\n- Following written orders, Greely’s party abandoned Fort Conger in August 1883 and traveled south. They eventually established a camp at Cape Sabine.\n- Food caches expected near Cape Sabine were inadequate. The party faced extreme hunger, cold, illness, isolation, and internal conflict.\n- Several earlier rescue attempts failed. In 1884, the U.S. Navy organized another expedition under Commander Winfield Scott Schley.\n- Schley’s squadron included the steam-powered ships Thetis, Bear, and Alert. Bear and Thetis reached Cape Sabine on June 22, 1884.\n- Rescuers found seven men alive; one of those seven, Joseph Elison, died during the return voyage. Of the original 25, six ultimately returned alive.\n- Recovery of scientific records was considered an important outcome of the expedition.\n- The expedition took place in an Arctic homeland that was already known, traveled, and inhabited by Indigenous peoples. Nineteenth-century U.S. accounts often described such regions as “unknown” or “unexplored,” language that reflected outsiders’ perspectives rather than an absence of existing knowledge.\n- The exhibition contains these four objects:\n  1. A brass-and-glass aneroid barometer used for field observation, maker unknown, circa 1880.\n  2. A water-stained page from a meteorological observation ledger kept at Fort Conger in February 1882; it records temperature, pressure, cloud cover, and wind.\n  3. A blue wool uniform coat attributed to Winfield Scott Schley and worn during the 1884 rescue expedition; attribution is supported by family provenance but is not conclusive.\n  4. An 1885 newspaper engraving depicting the survivors’ arrival in Portsmouth, New Hampshire. It was produced after the event for a mass audience and should be treated as a mediated representation, not an eyewitness record.\n\nAUDIENCE AND VOICE\n- Primary audience: visitors aged 14 and older with no prior knowledge of the event.\n- Reading level: approximately U.S. grade 8.\n- Tone: restrained, humane, vivid, and precise—not triumphalist, sensational, or melodramatic.\n- Center survival and scientific work without turning suffering into spectacle.\n- Do not frame the Arctic as empty, conquered, or waiting to be discovered.\n- Explain unfamiliar terms in context. Prefer short sentences and active voice.\n- Avoid rhetorical questions, exclamation points, invented scene-setting, and clichés such as “frozen wasteland,” “race against time,” or “against all odds.”\n- Refer to the group as “the Greely party,” not “Greely’s men,” except where grammar makes that impossible.\n\nDELIVERABLES\n1. Exhibition title and subtitle. Retain “North by Necessity” as the title and write a subtitle of no more than 12 words.\n2. Introductory panel: 170–190 words. Establish the scientific mission, failed relief efforts, move to Cape Sabine, rescue, human cost, and the need to question period ideas of “exploration.”\n3. Section panel, “Measuring the Arctic”: 105–125 words. Explain why sustained observations mattered and how routine recordkeeping continued under difficult conditions.\n4. Section panel, “Relief Delayed”: 105–125 words. Clarify that the disaster resulted from both environmental hazards and human decisions. Avoid assigning blame beyond the dossier.\n5. Section panel, “Rescue and Return”: 105–125 words. Describe Schley’s expedition, the June 22 rescue, the number of survivors, and the recovery of records without presenting the outcome as an uncomplicated victory.\n6. Four object labels, one for each listed object, each 65–80 words. Begin each label with a display heading of no more than six words. For the coat, clearly communicate the uncertain attribution. For the engraving, teach visitors why a published image is not neutral evidence.\n7. A family conversation prompt of 35–45 words that invites observation and discussion without using a rhetorical question. It should work beside the ledger and barometer.\n8. A 90–110 word accessibility-oriented visual description of the newspaper engraving. Because no image is provided, limit the description to what can responsibly be inferred from the dossier: medium, subject, publication context, and evidentiary limits. Explicitly state that composition and depicted details must be confirmed against the physical object before publication.\n9. A curator review table with columns: “Copy location,” “Claim or wording needing review,” and “Reason.” Flag at least four issues, including Indigenous context, the coat attribution, the engraving’s evidentiary status, and survivor totals.\n\nFORMAT\nPresent the finished package in the deliverable order above with clear headings. Put word counts in parentheses after every panel, label, prompt, and visual description. Do not include citations in visitor-facing copy. After the visitor-facing materials, include the curator review table and then a brief style-compliance checklist confirming that all required facts, cautions, and word limits have been addressed."
                }
            ]
        })


asyncio.run(main())

curl https://api.runware.ai/v1 \
  -H "Authorization: Bearer $RUNWARE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '[
    {
      "taskType": "textInference",
      "taskUUID": "78feeb31-af15-4993-82be-984eb368c5da",
      "model": "google:gemma@4-31b",
      "messages": [
        {
          "role": "user",
          "content": "Act as the lead interpretive writer for a maritime museum. Create the final gallery text package for “North by Necessity,” a small exhibition about the 1884 rescue of the Greely Arctic expedition. The writing must be historically responsible, engaging for general visitors, and ready for curatorial review.\n\nUse only the verified dossier below. Do not introduce quotations, motives, dates, Indigenous names, technical specifications, or other facts that are not supplied. Where the evidence does not support certainty, use appropriately qualified language.\n\nVERIFIED DOSSIER\n- In 1881, U.S. Army First Lieutenant Adolphus Greely led 25 men to Lady Franklin Bay on Ellesmere Island as part of the First International Polar Year.\n- The party established Fort Conger and conducted meteorological, astronomical, magnetic, tidal, and geographic observations.\n- In 1882 and 1883, scheduled relief ships failed to reach the party because of ice and poor decisions within the relief efforts.\n- Following written orders, Greely’s party abandoned Fort Conger in August 1883 and traveled south. They eventually established a camp at Cape Sabine.\n- Food caches expected near Cape Sabine were inadequate. The party faced extreme hunger, cold, illness, isolation, and internal conflict.\n- Several earlier rescue attempts failed. In 1884, the U.S. Navy organized another expedition under Commander Winfield Scott Schley.\n- Schley’s squadron included the steam-powered ships Thetis, Bear, and Alert. Bear and Thetis reached Cape Sabine on June 22, 1884.\n- Rescuers found seven men alive; one of those seven, Joseph Elison, died during the return voyage. Of the original 25, six ultimately returned alive.\n- Recovery of scientific records was considered an important outcome of the expedition.\n- The expedition took place in an Arctic homeland that was already known, traveled, and inhabited by Indigenous peoples. Nineteenth-century U.S. accounts often described such regions as “unknown” or “unexplored,” language that reflected outsiders’ perspectives rather than an absence of existing knowledge.\n- The exhibition contains these four objects:\n  1. A brass-and-glass aneroid barometer used for field observation, maker unknown, circa 1880.\n  2. A water-stained page from a meteorological observation ledger kept at Fort Conger in February 1882; it records temperature, pressure, cloud cover, and wind.\n  3. A blue wool uniform coat attributed to Winfield Scott Schley and worn during the 1884 rescue expedition; attribution is supported by family provenance but is not conclusive.\n  4. An 1885 newspaper engraving depicting the survivors’ arrival in Portsmouth, New Hampshire. It was produced after the event for a mass audience and should be treated as a mediated representation, not an eyewitness record.\n\nAUDIENCE AND VOICE\n- Primary audience: visitors aged 14 and older with no prior knowledge of the event.\n- Reading level: approximately U.S. grade 8.\n- Tone: restrained, humane, vivid, and precise—not triumphalist, sensational, or melodramatic.\n- Center survival and scientific work without turning suffering into spectacle.\n- Do not frame the Arctic as empty, conquered, or waiting to be discovered.\n- Explain unfamiliar terms in context. Prefer short sentences and active voice.\n- Avoid rhetorical questions, exclamation points, invented scene-setting, and clichés such as “frozen wasteland,” “race against time,” or “against all odds.”\n- Refer to the group as “the Greely party,” not “Greely’s men,” except where grammar makes that impossible.\n\nDELIVERABLES\n1. Exhibition title and subtitle. Retain “North by Necessity” as the title and write a subtitle of no more than 12 words.\n2. Introductory panel: 170–190 words. Establish the scientific mission, failed relief efforts, move to Cape Sabine, rescue, human cost, and the need to question period ideas of “exploration.”\n3. Section panel, “Measuring the Arctic”: 105–125 words. Explain why sustained observations mattered and how routine recordkeeping continued under difficult conditions.\n4. Section panel, “Relief Delayed”: 105–125 words. Clarify that the disaster resulted from both environmental hazards and human decisions. Avoid assigning blame beyond the dossier.\n5. Section panel, “Rescue and Return”: 105–125 words. Describe Schley’s expedition, the June 22 rescue, the number of survivors, and the recovery of records without presenting the outcome as an uncomplicated victory.\n6. Four object labels, one for each listed object, each 65–80 words. Begin each label with a display heading of no more than six words. For the coat, clearly communicate the uncertain attribution. For the engraving, teach visitors why a published image is not neutral evidence.\n7. A family conversation prompt of 35–45 words that invites observation and discussion without using a rhetorical question. It should work beside the ledger and barometer.\n8. A 90–110 word accessibility-oriented visual description of the newspaper engraving. Because no image is provided, limit the description to what can responsibly be inferred from the dossier: medium, subject, publication context, and evidentiary limits. Explicitly state that composition and depicted details must be confirmed against the physical object before publication.\n9. A curator review table with columns: “Copy location,” “Claim or wording needing review,” and “Reason.” Flag at least four issues, including Indigenous context, the coat attribution, the engraving’s evidentiary status, and survivor totals.\n\nFORMAT\nPresent the finished package in the deliverable order above with clear headings. Put word counts in parentheses after every panel, label, prompt, and visual description. Do not include citations in visitor-facing copy. After the visitor-facing materials, include the curator review table and then a brief style-compliance checklist confirming that all required facts, cautions, and word limits have been addressed."
        }
      ]
    }
  ]'

runware run google:gemma@4-31b \
  messages.0.role=user \
  messages.0.content="Act as the lead interpretive writer for a maritime museum. Create the final gallery text package for “North by Necessity,” a small exhibition about the 1884 rescue of the Greely Arctic expedition. The writing must be historically responsible, engaging for general visitors, and ready for curatorial review.

Use only the verified dossier below. Do not introduce quotations, motives, dates, Indigenous names, technical specifications, or other facts that are not supplied. Where the evidence does not support certainty, use appropriately qualified language.

VERIFIED DOSSIER
- In 1881, U.S. Army First Lieutenant Adolphus Greely led 25 men to Lady Franklin Bay on Ellesmere Island as part of the First International Polar Year.
- The party established Fort Conger and conducted meteorological, astronomical, magnetic, tidal, and geographic observations.
- In 1882 and 1883, scheduled relief ships failed to reach the party because of ice and poor decisions within the relief efforts.
- Following written orders, Greely’s party abandoned Fort Conger in August 1883 and traveled south. They eventually established a camp at Cape Sabine.
- Food caches expected near Cape Sabine were inadequate. The party faced extreme hunger, cold, illness, isolation, and internal conflict.
- Several earlier rescue attempts failed. In 1884, the U.S. Navy organized another expedition under Commander Winfield Scott Schley.
- Schley’s squadron included the steam-powered ships Thetis, Bear, and Alert. Bear and Thetis reached Cape Sabine on June 22, 1884.
- Rescuers found seven men alive; one of those seven, Joseph Elison, died during the return voyage. Of the original 25, six ultimately returned alive.
- Recovery of scientific records was considered an important outcome of the expedition.
- The expedition took place in an Arctic homeland that was already known, traveled, and inhabited by Indigenous peoples. Nineteenth-century U.S. accounts often described such regions as “unknown” or “unexplored,” language that reflected outsiders’ perspectives rather than an absence of existing knowledge.
- The exhibition contains these four objects:
  1. A brass-and-glass aneroid barometer used for field observation, maker unknown, circa 1880.
  2. A water-stained page from a meteorological observation ledger kept at Fort Conger in February 1882; it records temperature, pressure, cloud cover, and wind.
  3. A blue wool uniform coat attributed to Winfield Scott Schley and worn during the 1884 rescue expedition; attribution is supported by family provenance but is not conclusive.
  4. An 1885 newspaper engraving depicting the survivors’ arrival in Portsmouth, New Hampshire. It was produced after the event for a mass audience and should be treated as a mediated representation, not an eyewitness record.

AUDIENCE AND VOICE
- Primary audience: visitors aged 14 and older with no prior knowledge of the event.
- Reading level: approximately U.S. grade 8.
- Tone: restrained, humane, vivid, and precise—not triumphalist, sensational, or melodramatic.
- Center survival and scientific work without turning suffering into spectacle.
- Do not frame the Arctic as empty, conquered, or waiting to be discovered.
- Explain unfamiliar terms in context. Prefer short sentences and active voice.
- Avoid rhetorical questions, exclamation points, invented scene-setting, and clichés such as “frozen wasteland,” “race against time,” or “against all odds.”
- Refer to the group as “the Greely party,” not “Greely’s men,” except where grammar makes that impossible.

DELIVERABLES
1. Exhibition title and subtitle. Retain “North by Necessity” as the title and write a subtitle of no more than 12 words.
2. Introductory panel: 170–190 words. Establish the scientific mission, failed relief efforts, move to Cape Sabine, rescue, human cost, and the need to question period ideas of “exploration.”
3. Section panel, “Measuring the Arctic”: 105–125 words. Explain why sustained observations mattered and how routine recordkeeping continued under difficult conditions.
4. Section panel, “Relief Delayed”: 105–125 words. Clarify that the disaster resulted from both environmental hazards and human decisions. Avoid assigning blame beyond the dossier.
5. Section panel, “Rescue and Return”: 105–125 words. Describe Schley’s expedition, the June 22 rescue, the number of survivors, and the recovery of records without presenting the outcome as an uncomplicated victory.
6. Four object labels, one for each listed object, each 65–80 words. Begin each label with a display heading of no more than six words. For the coat, clearly communicate the uncertain attribution. For the engraving, teach visitors why a published image is not neutral evidence.
7. A family conversation prompt of 35–45 words that invites observation and discussion without using a rhetorical question. It should work beside the ledger and barometer.
8. A 90–110 word accessibility-oriented visual description of the newspaper engraving. Because no image is provided, limit the description to what can responsibly be inferred from the dossier: medium, subject, publication context, and evidentiary limits. Explicitly state that composition and depicted details must be confirmed against the physical object before publication.
9. A curator review table with columns: “Copy location,” “Claim or wording needing review,” and “Reason.” Flag at least four issues, including Indigenous context, the coat attribution, the engraving’s evidentiary status, and survivor totals.

FORMAT
Present the finished package in the deliverable order above with clear headings. Put word counts in parentheses after every panel, label, prompt, and visual description. Do not include citations in visitor-facing copy. After the visitor-facing materials, include the curator review table and then a brief style-compliance checklist confirming that all required facts, cautions, and word limits have been addressed."

{
  "taskType": "textInference",
  "taskUUID": "78feeb31-af15-4993-82be-984eb368c5da",
  "model": "google:gemma@4-31b",
  "messages": [
    {
      "role": "user",
      "content": "Act as the lead interpretive writer for a maritime museum. Create the final gallery text package for “North by Necessity,” a small exhibition about the 1884 rescue of the Greely Arctic expedition. The writing must be historically responsible, engaging for general visitors, and ready for curatorial review.\n\nUse only the verified dossier below. Do not introduce quotations, motives, dates, Indigenous names, technical specifications, or other facts that are not supplied. Where the evidence does not support certainty, use appropriately qualified language.\n\nVERIFIED DOSSIER\n- In 1881, U.S. Army First Lieutenant Adolphus Greely led 25 men to Lady Franklin Bay on Ellesmere Island as part of the First International Polar Year.\n- The party established Fort Conger and conducted meteorological, astronomical, magnetic, tidal, and geographic observations.\n- In 1882 and 1883, scheduled relief ships failed to reach the party because of ice and poor decisions within the relief efforts.\n- Following written orders, Greely’s party abandoned Fort Conger in August 1883 and traveled south. They eventually established a camp at Cape Sabine.\n- Food caches expected near Cape Sabine were inadequate. The party faced extreme hunger, cold, illness, isolation, and internal conflict.\n- Several earlier rescue attempts failed. In 1884, the U.S. Navy organized another expedition under Commander Winfield Scott Schley.\n- Schley’s squadron included the steam-powered ships Thetis, Bear, and Alert. Bear and Thetis reached Cape Sabine on June 22, 1884.\n- Rescuers found seven men alive; one of those seven, Joseph Elison, died during the return voyage. Of the original 25, six ultimately returned alive.\n- Recovery of scientific records was considered an important outcome of the expedition.\n- The expedition took place in an Arctic homeland that was already known, traveled, and inhabited by Indigenous peoples. Nineteenth-century U.S. accounts often described such regions as “unknown” or “unexplored,” language that reflected outsiders’ perspectives rather than an absence of existing knowledge.\n- The exhibition contains these four objects:\n  1. A brass-and-glass aneroid barometer used for field observation, maker unknown, circa 1880.\n  2. A water-stained page from a meteorological observation ledger kept at Fort Conger in February 1882; it records temperature, pressure, cloud cover, and wind.\n  3. A blue wool uniform coat attributed to Winfield Scott Schley and worn during the 1884 rescue expedition; attribution is supported by family provenance but is not conclusive.\n  4. An 1885 newspaper engraving depicting the survivors’ arrival in Portsmouth, New Hampshire. It was produced after the event for a mass audience and should be treated as a mediated representation, not an eyewitness record.\n\nAUDIENCE AND VOICE\n- Primary audience: visitors aged 14 and older with no prior knowledge of the event.\n- Reading level: approximately U.S. grade 8.\n- Tone: restrained, humane, vivid, and precise—not triumphalist, sensational, or melodramatic.\n- Center survival and scientific work without turning suffering into spectacle.\n- Do not frame the Arctic as empty, conquered, or waiting to be discovered.\n- Explain unfamiliar terms in context. Prefer short sentences and active voice.\n- Avoid rhetorical questions, exclamation points, invented scene-setting, and clichés such as “frozen wasteland,” “race against time,” or “against all odds.”\n- Refer to the group as “the Greely party,” not “Greely’s men,” except where grammar makes that impossible.\n\nDELIVERABLES\n1. Exhibition title and subtitle. Retain “North by Necessity” as the title and write a subtitle of no more than 12 words.\n2. Introductory panel: 170–190 words. Establish the scientific mission, failed relief efforts, move to Cape Sabine, rescue, human cost, and the need to question period ideas of “exploration.”\n3. Section panel, “Measuring the Arctic”: 105–125 words. Explain why sustained observations mattered and how routine recordkeeping continued under difficult conditions.\n4. Section panel, “Relief Delayed”: 105–125 words. Clarify that the disaster resulted from both environmental hazards and human decisions. Avoid assigning blame beyond the dossier.\n5. Section panel, “Rescue and Return”: 105–125 words. Describe Schley’s expedition, the June 22 rescue, the number of survivors, and the recovery of records without presenting the outcome as an uncomplicated victory.\n6. Four object labels, one for each listed object, each 65–80 words. Begin each label with a display heading of no more than six words. For the coat, clearly communicate the uncertain attribution. For the engraving, teach visitors why a published image is not neutral evidence.\n7. A family conversation prompt of 35–45 words that invites observation and discussion without using a rhetorical question. It should work beside the ledger and barometer.\n8. A 90–110 word accessibility-oriented visual description of the newspaper engraving. Because no image is provided, limit the description to what can responsibly be inferred from the dossier: medium, subject, publication context, and evidentiary limits. Explicitly state that composition and depicted details must be confirmed against the physical object before publication.\n9. A curator review table with columns: “Copy location,” “Claim or wording needing review,” and “Reason.” Flag at least four issues, including Indigenous context, the coat attribution, the engraving’s evidentiary status, and survivor totals.\n\nFORMAT\nPresent the finished package in the deliverable order above with clear headings. Put word counts in parentheses after every panel, label, prompt, and visual description. Do not include citations in visitor-facing copy. After the visitor-facing materials, include the curator review table and then a brief style-compliance checklist confirming that all required facts, cautions, and word limits have been addressed."
    }
  ]
}

import { createClient } from '@runware/sdk'

const client = await createClient({ apiKey: process.env.RUNWARE_API_KEY })
await client.connect()

const [result] = await client.run({
  model: 'google:gemma@4-31b',
  inputs: {
    images: [
      'https://assets.runware.ai/assets/inputs/ce20ce51-0ae3-4891-869a-6f30435958da.jpg'
    ],
    videos: [
      'https://assets.runware.ai/assets/inputs/f8c16baf-af89-4a0e-b90c-f400f8ac3acc.mp4'
    ]
  },
  messages: [
    {
      role: 'user',
      content: 'Act as a senior visual-merchandising auditor. The image is the approved autumn entrance-display reference; the video is a walkthrough of the installed display. Compare them and produce a practical compliance audit for the store manager.\n\nYour response must include:\n1. A concise executive summary and an overall compliance rating of Compliant, Partially Compliant, or Non-Compliant.\n2. A table with columns: Display Area, Approved Standard, Observed Installation, Status, Video Timestamp, and Recommended Correction.\n3. Separate findings for mannequin styling and order, plinth heights and spacing, central hiking-pack presentation, right-bay folded merchandise, color balance, signage, lighting, sightlines, and general presentation quality.\n4. A prioritized action list divided into Fix Before Opening, Fix Within 24 Hours, and Monitor.\n5. A short reinspection checklist a manager can use on a phone.\n\nUse the image and video as the only sources of truth. Cite timestamps for video-based observations and distinguish clearly between confirmed discrepancies, apparent discrepancies that need an on-site check, and elements that cannot be verified. Do not invent product names, measurements, policies, or missing visual details.'
    }
  ]
})

import asyncio
import os

from runware import Runware


async def main():
    async with Runware(api_key=os.environ["RUNWARE_API_KEY"]) as client:
        results = await client.run({
            "model": "google:gemma@4-31b",
            "inputs": {
                "images": [
                    "https://assets.runware.ai/assets/inputs/ce20ce51-0ae3-4891-869a-6f30435958da.jpg"
                ],
                "videos": [
                    "https://assets.runware.ai/assets/inputs/f8c16baf-af89-4a0e-b90c-f400f8ac3acc.mp4"
                ]
            },
            "messages": [
                {
                    "role": "user",
                    "content": "Act as a senior visual-merchandising auditor. The image is the approved autumn entrance-display reference; the video is a walkthrough of the installed display. Compare them and produce a practical compliance audit for the store manager.\n\nYour response must include:\n1. A concise executive summary and an overall compliance rating of Compliant, Partially Compliant, or Non-Compliant.\n2. A table with columns: Display Area, Approved Standard, Observed Installation, Status, Video Timestamp, and Recommended Correction.\n3. Separate findings for mannequin styling and order, plinth heights and spacing, central hiking-pack presentation, right-bay folded merchandise, color balance, signage, lighting, sightlines, and general presentation quality.\n4. A prioritized action list divided into Fix Before Opening, Fix Within 24 Hours, and Monitor.\n5. A short reinspection checklist a manager can use on a phone.\n\nUse the image and video as the only sources of truth. Cite timestamps for video-based observations and distinguish clearly between confirmed discrepancies, apparent discrepancies that need an on-site check, and elements that cannot be verified. Do not invent product names, measurements, policies, or missing visual details."
                }
            ]
        })


asyncio.run(main())

curl https://api.runware.ai/v1 \
  -H "Authorization: Bearer $RUNWARE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '[
    {
      "taskType": "textInference",
      "taskUUID": "b4138620-4483-414b-b6d1-b5b8959b54eb",
      "model": "google:gemma@4-31b",
      "inputs": {
        "images": [
          "https://assets.runware.ai/assets/inputs/ce20ce51-0ae3-4891-869a-6f30435958da.jpg"
        ],
        "videos": [
          "https://assets.runware.ai/assets/inputs/f8c16baf-af89-4a0e-b90c-f400f8ac3acc.mp4"
        ]
      },
      "messages": [
        {
          "role": "user",
          "content": "Act as a senior visual-merchandising auditor. The image is the approved autumn entrance-display reference; the video is a walkthrough of the installed display. Compare them and produce a practical compliance audit for the store manager.\n\nYour response must include:\n1. A concise executive summary and an overall compliance rating of Compliant, Partially Compliant, or Non-Compliant.\n2. A table with columns: Display Area, Approved Standard, Observed Installation, Status, Video Timestamp, and Recommended Correction.\n3. Separate findings for mannequin styling and order, plinth heights and spacing, central hiking-pack presentation, right-bay folded merchandise, color balance, signage, lighting, sightlines, and general presentation quality.\n4. A prioritized action list divided into Fix Before Opening, Fix Within 24 Hours, and Monitor.\n5. A short reinspection checklist a manager can use on a phone.\n\nUse the image and video as the only sources of truth. Cite timestamps for video-based observations and distinguish clearly between confirmed discrepancies, apparent discrepancies that need an on-site check, and elements that cannot be verified. Do not invent product names, measurements, policies, or missing visual details."
        }
      ]
    }
  ]'

runware run google:gemma@4-31b \
  inputs.images.0=https://assets.runware.ai/assets/inputs/ce20ce51-0ae3-4891-869a-6f30435958da.jpg \
  inputs.videos.0=https://assets.runware.ai/assets/inputs/f8c16baf-af89-4a0e-b90c-f400f8ac3acc.mp4 \
  messages.0.role=user \
  messages.0.content="Act as a senior visual-merchandising auditor. The image is the approved autumn entrance-display reference; the video is a walkthrough of the installed display. Compare them and produce a practical compliance audit for the store manager.

Your response must include:
1. A concise executive summary and an overall compliance rating of Compliant, Partially Compliant, or Non-Compliant.
2. A table with columns: Display Area, Approved Standard, Observed Installation, Status, Video Timestamp, and Recommended Correction.
3. Separate findings for mannequin styling and order, plinth heights and spacing, central hiking-pack presentation, right-bay folded merchandise, color balance, signage, lighting, sightlines, and general presentation quality.
4. A prioritized action list divided into Fix Before Opening, Fix Within 24 Hours, and Monitor.
5. A short reinspection checklist a manager can use on a phone.

Use the image and video as the only sources of truth. Cite timestamps for video-based observations and distinguish clearly between confirmed discrepancies, apparent discrepancies that need an on-site check, and elements that cannot be verified. Do not invent product names, measurements, policies, or missing visual details."

{
  "taskType": "textInference",
  "taskUUID": "b4138620-4483-414b-b6d1-b5b8959b54eb",
  "model": "google:gemma@4-31b",
  "inputs": {
    "images": [
      "https://assets.runware.ai/assets/inputs/ce20ce51-0ae3-4891-869a-6f30435958da.jpg"
    ],
    "videos": [
      "https://assets.runware.ai/assets/inputs/f8c16baf-af89-4a0e-b90c-f400f8ac3acc.mp4"
    ]
  },
  "messages": [
    {
      "role": "user",
      "content": "Act as a senior visual-merchandising auditor. The image is the approved autumn entrance-display reference; the video is a walkthrough of the installed display. Compare them and produce a practical compliance audit for the store manager.\n\nYour response must include:\n1. A concise executive summary and an overall compliance rating of Compliant, Partially Compliant, or Non-Compliant.\n2. A table with columns: Display Area, Approved Standard, Observed Installation, Status, Video Timestamp, and Recommended Correction.\n3. Separate findings for mannequin styling and order, plinth heights and spacing, central hiking-pack presentation, right-bay folded merchandise, color balance, signage, lighting, sightlines, and general presentation quality.\n4. A prioritized action list divided into Fix Before Opening, Fix Within 24 Hours, and Monitor.\n5. A short reinspection checklist a manager can use on a phone.\n\nUse the image and video as the only sources of truth. Cite timestamps for video-based observations and distinguish clearly between confirmed discrepancies, apparent discrepancies that need an on-site check, and elements that cannot be verified. Do not invent product names, measurements, policies, or missing visual details."
    }
  ]
}

import { createClient } from '@runware/sdk'

const client = await createClient({ apiKey: process.env.RUNWARE_API_KEY })
await client.connect()

const [result] = await client.run({
  model: 'google:gemma@4-31b',
  messages: [
    {
      role: 'user',
      content: 'Act as an energy-infrastructure strategy consultant. Write a board-ready investment memo for Northline Cold Logistics, a regional seafood cold-storage operator, recommending whether to procure a microgrid for its Harbor Point facility. Use only the facts and assumptions below; do not introduce outside data.\n\nAUDIENCE AND DECISION\nThe memo will go to the CEO, CFO, COO, and two nontechnical board members. They must decide whether to select one of three binding vendor proposals or defer the project for one year. Their priorities, in order, are: preventing inventory loss, maintaining an acceptable financial return, limiting implementation risk, and reducing emissions.\n\nFACILITY BASELINE\n- Harbor Point is a 24/7 refrigerated warehouse in coastal Maine.\n- Annual electricity consumption: 8,400 MWh.\n- Peak demand: 1.65 MW.\n- Current blended electricity cost: $0.142/kWh, comprising energy, demand, and delivery charges.\n- Assume the blended cost rises 3% annually.\n- A controlled shutdown requires four hours; an uncontrolled outage begins threatening product integrity after six hours.\n- Average inventory value on site: $11.8 million.\n- Management estimates a 35% probability of at least one outage lasting longer than six hours during any five-year period.\n- Expected uninsured loss from such an event, after salvage and customer recoveries: $2.4 million.\n- For this analysis, annualize the outage risk as 7% per year and treat avoided expected loss as 7% × $2.4 million × the proposal’s stated share of long-outage exposure mitigated.\n- Existing diesel backup supports only safety systems and office IT, not refrigeration.\n- Current operational emissions: 2,950 metric tons CO2e per year.\n\nFINANCIAL RULES\n- Evaluation period: 15 years from commercial operation.\n- Discount rate: 8% nominal.\n- All listed annual benefits and operating costs begin at the end of Year 1.\n- Electricity savings increase 3% annually unless otherwise stated.\n- O&M costs increase 2% annually.\n- Carbon has no cash value in the base case.\n- Ignore taxes, depreciation, financing costs, residual value, and construction-period discounting.\n- Treat grants as reductions to upfront capital cost.\n- Battery replacement costs occur at the end of the specified year.\n- Calculate net present value, simple payback, and approximate internal rate of return for each proposal. Round dollar results to the nearest $10,000 and percentages to one decimal place. Clearly label reasonable approximations.\n\nPROPOSAL A — TIDAL GRIDWORKS\n- 1.2 MW rooftop solar plus 3.6 MWh lithium-ion battery.\n- Gross capital cost: $5.60 million.\n- Confirmed state resilience grant: $900,000.\n- Annual Year 1 electricity savings: $545,000.\n- Annual Year 1 O&M: $74,000.\n- Battery replacement: $1.05 million at end of Year 10.\n- Mitigates 78% of modeled long-outage exposure.\n- Estimated emissions reduction: 1,090 metric tons CO2e annually.\n- Can sustain critical refrigeration for approximately 10 hours without grid power under typical conditions.\n- Fixed-price EPC contract; completion in 11 months.\n- Vendor has completed nine comparable cold-storage projects.\n- Roof reinforcement is included, but final structural verification remains a condition before notice to proceed.\n\nPROPOSAL B — PINE STATE ENERGY\n- 1.0 MW solar, 2.4 MWh battery, and 1.1 MW renewable-diesel generator.\n- Gross capital cost: $6.35 million.\n- Confirmed federal grant: $1.25 million.\n- Annual Year 1 electricity savings: $495,000.\n- Annual Year 1 O&M: $128,000.\n- Battery replacement: $720,000 at end of Year 10.\n- Mitigates 96% of modeled long-outage exposure, assuming fuel delivery remains available.\n- Estimated emissions reduction: 820 metric tons CO2e annually under expected generator use.\n- Can sustain critical refrigeration for at least 48 hours with a full fuel tank.\n- Completion in 14 months.\n- Vendor has completed three comparable cold-storage projects and 22 hospital microgrids.\n- Contract price may increase by up to $300,000 if interconnection protection upgrades are required; assign a 40% probability to this cost and include its expected value in the base-case economics.\n\nPROPOSAL C — ATLANTIC STORAGE SYSTEMS\n- 4.8 MWh battery with no on-site generation.\n- Gross capital cost: $3.45 million.\n- No grant.\n- Annual Year 1 electricity savings: $430,000.\n- Annual Year 1 O&M: $52,000.\n- Battery replacement: $1.30 million at end of Year 9.\n- Mitigates 52% of modeled long-outage exposure.\n- Estimated emissions reduction: 210 metric tons CO2e annually.\n- Can sustain critical refrigeration for approximately eight hours at current loads.\n- Completion in seven months.\n- Vendor has completed 31 commercial battery projects, but none at refrigerated warehouses.\n- The proposal guarantees 90% usable battery capacity through Year 8; there is no performance guarantee after replacement.\n\nDEFERRAL OPTION\n- No Year 0 capital outlay.\n- The procurement adviser expects equipment and labor pricing to rise 6% over the next year.\n- Tidal Gridworks’ state grant is reserved for nine months and would probably be lost if the decision is deferred.\n- Pine State Energy’s federal grant remains available for 18 months.\n- No vendor will hold its current completion schedule.\n- Outage exposure and current electricity costs continue during the delay.\n\nREQUIRED DELIVERABLE\nWrite a polished memo of approximately 1,500–1,900 words with these sections:\n1. To / From / Date / Subject header.\n2. Executive recommendation stating the decision in the first paragraph.\n3. Decision rationale, explaining why the recommended option best fits the board’s ordered priorities.\n4. A compact comparison table covering net upfront cost, Year 1 net operating benefit, annualized avoided outage loss, base-case NPV, simple payback, approximate IRR, outage capability, emissions reduction, schedule, and major execution risk.\n5. Financial analysis describing the cash-flow method and separating electricity savings, O&M, avoided expected inventory loss, replacement costs, and proposal-specific expected costs. Define simple payback as the first year cumulative undiscounted net cash flow becomes positive.\n6. Sensitivity analysis for the recommended proposal under three separate downside cases: electricity savings are 20% below forecast; no avoided-loss benefit is credited; and upfront cost is 10% higher. Report the revised NPV for each case and explain what each reveals.\n7. Key risks and mitigations, with named owners by role, such as COO, CFO, facilities director, or general counsel.\n8. A 90-day action plan with concrete milestones.\n9. A final board resolution written as one sentence that can be copied into meeting minutes.\n\nMaintain a decisive, sober tone. Explain technical concepts in plain English without oversimplifying the economics. Distinguish calculated facts from judgment. Do not recommend deferral merely because information is incomplete; assess whether the identified uncertainties can be managed through contract terms, diligence, contingencies, or approval conditions.'
    }
  ]
})

import asyncio
import os

from runware import Runware


async def main():
    async with Runware(api_key=os.environ["RUNWARE_API_KEY"]) as client:
        results = await client.run({
            "model": "google:gemma@4-31b",
            "messages": [
                {
                    "role": "user",
                    "content": "Act as an energy-infrastructure strategy consultant. Write a board-ready investment memo for Northline Cold Logistics, a regional seafood cold-storage operator, recommending whether to procure a microgrid for its Harbor Point facility. Use only the facts and assumptions below; do not introduce outside data.\n\nAUDIENCE AND DECISION\nThe memo will go to the CEO, CFO, COO, and two nontechnical board members. They must decide whether to select one of three binding vendor proposals or defer the project for one year. Their priorities, in order, are: preventing inventory loss, maintaining an acceptable financial return, limiting implementation risk, and reducing emissions.\n\nFACILITY BASELINE\n- Harbor Point is a 24/7 refrigerated warehouse in coastal Maine.\n- Annual electricity consumption: 8,400 MWh.\n- Peak demand: 1.65 MW.\n- Current blended electricity cost: $0.142/kWh, comprising energy, demand, and delivery charges.\n- Assume the blended cost rises 3% annually.\n- A controlled shutdown requires four hours; an uncontrolled outage begins threatening product integrity after six hours.\n- Average inventory value on site: $11.8 million.\n- Management estimates a 35% probability of at least one outage lasting longer than six hours during any five-year period.\n- Expected uninsured loss from such an event, after salvage and customer recoveries: $2.4 million.\n- For this analysis, annualize the outage risk as 7% per year and treat avoided expected loss as 7% × $2.4 million × the proposal’s stated share of long-outage exposure mitigated.\n- Existing diesel backup supports only safety systems and office IT, not refrigeration.\n- Current operational emissions: 2,950 metric tons CO2e per year.\n\nFINANCIAL RULES\n- Evaluation period: 15 years from commercial operation.\n- Discount rate: 8% nominal.\n- All listed annual benefits and operating costs begin at the end of Year 1.\n- Electricity savings increase 3% annually unless otherwise stated.\n- O&M costs increase 2% annually.\n- Carbon has no cash value in the base case.\n- Ignore taxes, depreciation, financing costs, residual value, and construction-period discounting.\n- Treat grants as reductions to upfront capital cost.\n- Battery replacement costs occur at the end of the specified year.\n- Calculate net present value, simple payback, and approximate internal rate of return for each proposal. Round dollar results to the nearest $10,000 and percentages to one decimal place. Clearly label reasonable approximations.\n\nPROPOSAL A — TIDAL GRIDWORKS\n- 1.2 MW rooftop solar plus 3.6 MWh lithium-ion battery.\n- Gross capital cost: $5.60 million.\n- Confirmed state resilience grant: $900,000.\n- Annual Year 1 electricity savings: $545,000.\n- Annual Year 1 O&M: $74,000.\n- Battery replacement: $1.05 million at end of Year 10.\n- Mitigates 78% of modeled long-outage exposure.\n- Estimated emissions reduction: 1,090 metric tons CO2e annually.\n- Can sustain critical refrigeration for approximately 10 hours without grid power under typical conditions.\n- Fixed-price EPC contract; completion in 11 months.\n- Vendor has completed nine comparable cold-storage projects.\n- Roof reinforcement is included, but final structural verification remains a condition before notice to proceed.\n\nPROPOSAL B — PINE STATE ENERGY\n- 1.0 MW solar, 2.4 MWh battery, and 1.1 MW renewable-diesel generator.\n- Gross capital cost: $6.35 million.\n- Confirmed federal grant: $1.25 million.\n- Annual Year 1 electricity savings: $495,000.\n- Annual Year 1 O&M: $128,000.\n- Battery replacement: $720,000 at end of Year 10.\n- Mitigates 96% of modeled long-outage exposure, assuming fuel delivery remains available.\n- Estimated emissions reduction: 820 metric tons CO2e annually under expected generator use.\n- Can sustain critical refrigeration for at least 48 hours with a full fuel tank.\n- Completion in 14 months.\n- Vendor has completed three comparable cold-storage projects and 22 hospital microgrids.\n- Contract price may increase by up to $300,000 if interconnection protection upgrades are required; assign a 40% probability to this cost and include its expected value in the base-case economics.\n\nPROPOSAL C — ATLANTIC STORAGE SYSTEMS\n- 4.8 MWh battery with no on-site generation.\n- Gross capital cost: $3.45 million.\n- No grant.\n- Annual Year 1 electricity savings: $430,000.\n- Annual Year 1 O&M: $52,000.\n- Battery replacement: $1.30 million at end of Year 9.\n- Mitigates 52% of modeled long-outage exposure.\n- Estimated emissions reduction: 210 metric tons CO2e annually.\n- Can sustain critical refrigeration for approximately eight hours at current loads.\n- Completion in seven months.\n- Vendor has completed 31 commercial battery projects, but none at refrigerated warehouses.\n- The proposal guarantees 90% usable battery capacity through Year 8; there is no performance guarantee after replacement.\n\nDEFERRAL OPTION\n- No Year 0 capital outlay.\n- The procurement adviser expects equipment and labor pricing to rise 6% over the next year.\n- Tidal Gridworks’ state grant is reserved for nine months and would probably be lost if the decision is deferred.\n- Pine State Energy’s federal grant remains available for 18 months.\n- No vendor will hold its current completion schedule.\n- Outage exposure and current electricity costs continue during the delay.\n\nREQUIRED DELIVERABLE\nWrite a polished memo of approximately 1,500–1,900 words with these sections:\n1. To / From / Date / Subject header.\n2. Executive recommendation stating the decision in the first paragraph.\n3. Decision rationale, explaining why the recommended option best fits the board’s ordered priorities.\n4. A compact comparison table covering net upfront cost, Year 1 net operating benefit, annualized avoided outage loss, base-case NPV, simple payback, approximate IRR, outage capability, emissions reduction, schedule, and major execution risk.\n5. Financial analysis describing the cash-flow method and separating electricity savings, O&M, avoided expected inventory loss, replacement costs, and proposal-specific expected costs. Define simple payback as the first year cumulative undiscounted net cash flow becomes positive.\n6. Sensitivity analysis for the recommended proposal under three separate downside cases: electricity savings are 20% below forecast; no avoided-loss benefit is credited; and upfront cost is 10% higher. Report the revised NPV for each case and explain what each reveals.\n7. Key risks and mitigations, with named owners by role, such as COO, CFO, facilities director, or general counsel.\n8. A 90-day action plan with concrete milestones.\n9. A final board resolution written as one sentence that can be copied into meeting minutes.\n\nMaintain a decisive, sober tone. Explain technical concepts in plain English without oversimplifying the economics. Distinguish calculated facts from judgment. Do not recommend deferral merely because information is incomplete; assess whether the identified uncertainties can be managed through contract terms, diligence, contingencies, or approval conditions."
                }
            ]
        })


asyncio.run(main())

curl https://api.runware.ai/v1 \
  -H "Authorization: Bearer $RUNWARE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '[
    {
      "taskType": "textInference",
      "taskUUID": "67646ffa-2dff-4dd7-8b8d-8625d73bbe6e",
      "model": "google:gemma@4-31b",
      "messages": [
        {
          "role": "user",
          "content": "Act as an energy-infrastructure strategy consultant. Write a board-ready investment memo for Northline Cold Logistics, a regional seafood cold-storage operator, recommending whether to procure a microgrid for its Harbor Point facility. Use only the facts and assumptions below; do not introduce outside data.\n\nAUDIENCE AND DECISION\nThe memo will go to the CEO, CFO, COO, and two nontechnical board members. They must decide whether to select one of three binding vendor proposals or defer the project for one year. Their priorities, in order, are: preventing inventory loss, maintaining an acceptable financial return, limiting implementation risk, and reducing emissions.\n\nFACILITY BASELINE\n- Harbor Point is a 24/7 refrigerated warehouse in coastal Maine.\n- Annual electricity consumption: 8,400 MWh.\n- Peak demand: 1.65 MW.\n- Current blended electricity cost: $0.142/kWh, comprising energy, demand, and delivery charges.\n- Assume the blended cost rises 3% annually.\n- A controlled shutdown requires four hours; an uncontrolled outage begins threatening product integrity after six hours.\n- Average inventory value on site: $11.8 million.\n- Management estimates a 35% probability of at least one outage lasting longer than six hours during any five-year period.\n- Expected uninsured loss from such an event, after salvage and customer recoveries: $2.4 million.\n- For this analysis, annualize the outage risk as 7% per year and treat avoided expected loss as 7% × $2.4 million × the proposal’s stated share of long-outage exposure mitigated.\n- Existing diesel backup supports only safety systems and office IT, not refrigeration.\n- Current operational emissions: 2,950 metric tons CO2e per year.\n\nFINANCIAL RULES\n- Evaluation period: 15 years from commercial operation.\n- Discount rate: 8% nominal.\n- All listed annual benefits and operating costs begin at the end of Year 1.\n- Electricity savings increase 3% annually unless otherwise stated.\n- O&M costs increase 2% annually.\n- Carbon has no cash value in the base case.\n- Ignore taxes, depreciation, financing costs, residual value, and construction-period discounting.\n- Treat grants as reductions to upfront capital cost.\n- Battery replacement costs occur at the end of the specified year.\n- Calculate net present value, simple payback, and approximate internal rate of return for each proposal. Round dollar results to the nearest $10,000 and percentages to one decimal place. Clearly label reasonable approximations.\n\nPROPOSAL A — TIDAL GRIDWORKS\n- 1.2 MW rooftop solar plus 3.6 MWh lithium-ion battery.\n- Gross capital cost: $5.60 million.\n- Confirmed state resilience grant: $900,000.\n- Annual Year 1 electricity savings: $545,000.\n- Annual Year 1 O&M: $74,000.\n- Battery replacement: $1.05 million at end of Year 10.\n- Mitigates 78% of modeled long-outage exposure.\n- Estimated emissions reduction: 1,090 metric tons CO2e annually.\n- Can sustain critical refrigeration for approximately 10 hours without grid power under typical conditions.\n- Fixed-price EPC contract; completion in 11 months.\n- Vendor has completed nine comparable cold-storage projects.\n- Roof reinforcement is included, but final structural verification remains a condition before notice to proceed.\n\nPROPOSAL B — PINE STATE ENERGY\n- 1.0 MW solar, 2.4 MWh battery, and 1.1 MW renewable-diesel generator.\n- Gross capital cost: $6.35 million.\n- Confirmed federal grant: $1.25 million.\n- Annual Year 1 electricity savings: $495,000.\n- Annual Year 1 O&M: $128,000.\n- Battery replacement: $720,000 at end of Year 10.\n- Mitigates 96% of modeled long-outage exposure, assuming fuel delivery remains available.\n- Estimated emissions reduction: 820 metric tons CO2e annually under expected generator use.\n- Can sustain critical refrigeration for at least 48 hours with a full fuel tank.\n- Completion in 14 months.\n- Vendor has completed three comparable cold-storage projects and 22 hospital microgrids.\n- Contract price may increase by up to $300,000 if interconnection protection upgrades are required; assign a 40% probability to this cost and include its expected value in the base-case economics.\n\nPROPOSAL C — ATLANTIC STORAGE SYSTEMS\n- 4.8 MWh battery with no on-site generation.\n- Gross capital cost: $3.45 million.\n- No grant.\n- Annual Year 1 electricity savings: $430,000.\n- Annual Year 1 O&M: $52,000.\n- Battery replacement: $1.30 million at end of Year 9.\n- Mitigates 52% of modeled long-outage exposure.\n- Estimated emissions reduction: 210 metric tons CO2e annually.\n- Can sustain critical refrigeration for approximately eight hours at current loads.\n- Completion in seven months.\n- Vendor has completed 31 commercial battery projects, but none at refrigerated warehouses.\n- The proposal guarantees 90% usable battery capacity through Year 8; there is no performance guarantee after replacement.\n\nDEFERRAL OPTION\n- No Year 0 capital outlay.\n- The procurement adviser expects equipment and labor pricing to rise 6% over the next year.\n- Tidal Gridworks’ state grant is reserved for nine months and would probably be lost if the decision is deferred.\n- Pine State Energy’s federal grant remains available for 18 months.\n- No vendor will hold its current completion schedule.\n- Outage exposure and current electricity costs continue during the delay.\n\nREQUIRED DELIVERABLE\nWrite a polished memo of approximately 1,500–1,900 words with these sections:\n1. To / From / Date / Subject header.\n2. Executive recommendation stating the decision in the first paragraph.\n3. Decision rationale, explaining why the recommended option best fits the board’s ordered priorities.\n4. A compact comparison table covering net upfront cost, Year 1 net operating benefit, annualized avoided outage loss, base-case NPV, simple payback, approximate IRR, outage capability, emissions reduction, schedule, and major execution risk.\n5. Financial analysis describing the cash-flow method and separating electricity savings, O&M, avoided expected inventory loss, replacement costs, and proposal-specific expected costs. Define simple payback as the first year cumulative undiscounted net cash flow becomes positive.\n6. Sensitivity analysis for the recommended proposal under three separate downside cases: electricity savings are 20% below forecast; no avoided-loss benefit is credited; and upfront cost is 10% higher. Report the revised NPV for each case and explain what each reveals.\n7. Key risks and mitigations, with named owners by role, such as COO, CFO, facilities director, or general counsel.\n8. A 90-day action plan with concrete milestones.\n9. A final board resolution written as one sentence that can be copied into meeting minutes.\n\nMaintain a decisive, sober tone. Explain technical concepts in plain English without oversimplifying the economics. Distinguish calculated facts from judgment. Do not recommend deferral merely because information is incomplete; assess whether the identified uncertainties can be managed through contract terms, diligence, contingencies, or approval conditions."
        }
      ]
    }
  ]'

runware run google:gemma@4-31b \
  messages.0.role=user \
  messages.0.content="Act as an energy-infrastructure strategy consultant. Write a board-ready investment memo for Northline Cold Logistics, a regional seafood cold-storage operator, recommending whether to procure a microgrid for its Harbor Point facility. Use only the facts and assumptions below; do not introduce outside data.

AUDIENCE AND DECISION
The memo will go to the CEO, CFO, COO, and two nontechnical board members. They must decide whether to select one of three binding vendor proposals or defer the project for one year. Their priorities, in order, are: preventing inventory loss, maintaining an acceptable financial return, limiting implementation risk, and reducing emissions.

FACILITY BASELINE
- Harbor Point is a 24/7 refrigerated warehouse in coastal Maine.
- Annual electricity consumption: 8,400 MWh.
- Peak demand: 1.65 MW.
- Current blended electricity cost: \$0.142/kWh, comprising energy, demand, and delivery charges.
- Assume the blended cost rises 3% annually.
- A controlled shutdown requires four hours; an uncontrolled outage begins threatening product integrity after six hours.
- Average inventory value on site: \$11.8 million.
- Management estimates a 35% probability of at least one outage lasting longer than six hours during any five-year period.
- Expected uninsured loss from such an event, after salvage and customer recoveries: \$2.4 million.
- For this analysis, annualize the outage risk as 7% per year and treat avoided expected loss as 7% × \$2.4 million × the proposal’s stated share of long-outage exposure mitigated.
- Existing diesel backup supports only safety systems and office IT, not refrigeration.
- Current operational emissions: 2,950 metric tons CO2e per year.

FINANCIAL RULES
- Evaluation period: 15 years from commercial operation.
- Discount rate: 8% nominal.
- All listed annual benefits and operating costs begin at the end of Year 1.
- Electricity savings increase 3% annually unless otherwise stated.
- O&M costs increase 2% annually.
- Carbon has no cash value in the base case.
- Ignore taxes, depreciation, financing costs, residual value, and construction-period discounting.
- Treat grants as reductions to upfront capital cost.
- Battery replacement costs occur at the end of the specified year.
- Calculate net present value, simple payback, and approximate internal rate of return for each proposal. Round dollar results to the nearest \$10,000 and percentages to one decimal place. Clearly label reasonable approximations.

PROPOSAL A — TIDAL GRIDWORKS
- 1.2 MW rooftop solar plus 3.6 MWh lithium-ion battery.
- Gross capital cost: \$5.60 million.
- Confirmed state resilience grant: \$900,000.
- Annual Year 1 electricity savings: \$545,000.
- Annual Year 1 O&M: \$74,000.
- Battery replacement: \$1.05 million at end of Year 10.
- Mitigates 78% of modeled long-outage exposure.
- Estimated emissions reduction: 1,090 metric tons CO2e annually.
- Can sustain critical refrigeration for approximately 10 hours without grid power under typical conditions.
- Fixed-price EPC contract; completion in 11 months.
- Vendor has completed nine comparable cold-storage projects.
- Roof reinforcement is included, but final structural verification remains a condition before notice to proceed.

PROPOSAL B — PINE STATE ENERGY
- 1.0 MW solar, 2.4 MWh battery, and 1.1 MW renewable-diesel generator.
- Gross capital cost: \$6.35 million.
- Confirmed federal grant: \$1.25 million.
- Annual Year 1 electricity savings: \$495,000.
- Annual Year 1 O&M: \$128,000.
- Battery replacement: \$720,000 at end of Year 10.
- Mitigates 96% of modeled long-outage exposure, assuming fuel delivery remains available.
- Estimated emissions reduction: 820 metric tons CO2e annually under expected generator use.
- Can sustain critical refrigeration for at least 48 hours with a full fuel tank.
- Completion in 14 months.
- Vendor has completed three comparable cold-storage projects and 22 hospital microgrids.
- Contract price may increase by up to \$300,000 if interconnection protection upgrades are required; assign a 40% probability to this cost and include its expected value in the base-case economics.

PROPOSAL C — ATLANTIC STORAGE SYSTEMS
- 4.8 MWh battery with no on-site generation.
- Gross capital cost: \$3.45 million.
- No grant.
- Annual Year 1 electricity savings: \$430,000.
- Annual Year 1 O&M: \$52,000.
- Battery replacement: \$1.30 million at end of Year 9.
- Mitigates 52% of modeled long-outage exposure.
- Estimated emissions reduction: 210 metric tons CO2e annually.
- Can sustain critical refrigeration for approximately eight hours at current loads.
- Completion in seven months.
- Vendor has completed 31 commercial battery projects, but none at refrigerated warehouses.
- The proposal guarantees 90% usable battery capacity through Year 8; there is no performance guarantee after replacement.

DEFERRAL OPTION
- No Year 0 capital outlay.
- The procurement adviser expects equipment and labor pricing to rise 6% over the next year.
- Tidal Gridworks’ state grant is reserved for nine months and would probably be lost if the decision is deferred.
- Pine State Energy’s federal grant remains available for 18 months.
- No vendor will hold its current completion schedule.
- Outage exposure and current electricity costs continue during the delay.

REQUIRED DELIVERABLE
Write a polished memo of approximately 1,500–1,900 words with these sections:
1. To / From / Date / Subject header.
2. Executive recommendation stating the decision in the first paragraph.
3. Decision rationale, explaining why the recommended option best fits the board’s ordered priorities.
4. A compact comparison table covering net upfront cost, Year 1 net operating benefit, annualized avoided outage loss, base-case NPV, simple payback, approximate IRR, outage capability, emissions reduction, schedule, and major execution risk.
5. Financial analysis describing the cash-flow method and separating electricity savings, O&M, avoided expected inventory loss, replacement costs, and proposal-specific expected costs. Define simple payback as the first year cumulative undiscounted net cash flow becomes positive.
6. Sensitivity analysis for the recommended proposal under three separate downside cases: electricity savings are 20% below forecast; no avoided-loss benefit is credited; and upfront cost is 10% higher. Report the revised NPV for each case and explain what each reveals.
7. Key risks and mitigations, with named owners by role, such as COO, CFO, facilities director, or general counsel.
8. A 90-day action plan with concrete milestones.
9. A final board resolution written as one sentence that can be copied into meeting minutes.

Maintain a decisive, sober tone. Explain technical concepts in plain English without oversimplifying the economics. Distinguish calculated facts from judgment. Do not recommend deferral merely because information is incomplete; assess whether the identified uncertainties can be managed through contract terms, diligence, contingencies, or approval conditions."

{
  "taskType": "textInference",
  "taskUUID": "67646ffa-2dff-4dd7-8b8d-8625d73bbe6e",
  "model": "google:gemma@4-31b",
  "messages": [
    {
      "role": "user",
      "content": "Act as an energy-infrastructure strategy consultant. Write a board-ready investment memo for Northline Cold Logistics, a regional seafood cold-storage operator, recommending whether to procure a microgrid for its Harbor Point facility. Use only the facts and assumptions below; do not introduce outside data.\n\nAUDIENCE AND DECISION\nThe memo will go to the CEO, CFO, COO, and two nontechnical board members. They must decide whether to select one of three binding vendor proposals or defer the project for one year. Their priorities, in order, are: preventing inventory loss, maintaining an acceptable financial return, limiting implementation risk, and reducing emissions.\n\nFACILITY BASELINE\n- Harbor Point is a 24/7 refrigerated warehouse in coastal Maine.\n- Annual electricity consumption: 8,400 MWh.\n- Peak demand: 1.65 MW.\n- Current blended electricity cost: $0.142/kWh, comprising energy, demand, and delivery charges.\n- Assume the blended cost rises 3% annually.\n- A controlled shutdown requires four hours; an uncontrolled outage begins threatening product integrity after six hours.\n- Average inventory value on site: $11.8 million.\n- Management estimates a 35% probability of at least one outage lasting longer than six hours during any five-year period.\n- Expected uninsured loss from such an event, after salvage and customer recoveries: $2.4 million.\n- For this analysis, annualize the outage risk as 7% per year and treat avoided expected loss as 7% × $2.4 million × the proposal’s stated share of long-outage exposure mitigated.\n- Existing diesel backup supports only safety systems and office IT, not refrigeration.\n- Current operational emissions: 2,950 metric tons CO2e per year.\n\nFINANCIAL RULES\n- Evaluation period: 15 years from commercial operation.\n- Discount rate: 8% nominal.\n- All listed annual benefits and operating costs begin at the end of Year 1.\n- Electricity savings increase 3% annually unless otherwise stated.\n- O&M costs increase 2% annually.\n- Carbon has no cash value in the base case.\n- Ignore taxes, depreciation, financing costs, residual value, and construction-period discounting.\n- Treat grants as reductions to upfront capital cost.\n- Battery replacement costs occur at the end of the specified year.\n- Calculate net present value, simple payback, and approximate internal rate of return for each proposal. Round dollar results to the nearest $10,000 and percentages to one decimal place. Clearly label reasonable approximations.\n\nPROPOSAL A — TIDAL GRIDWORKS\n- 1.2 MW rooftop solar plus 3.6 MWh lithium-ion battery.\n- Gross capital cost: $5.60 million.\n- Confirmed state resilience grant: $900,000.\n- Annual Year 1 electricity savings: $545,000.\n- Annual Year 1 O&M: $74,000.\n- Battery replacement: $1.05 million at end of Year 10.\n- Mitigates 78% of modeled long-outage exposure.\n- Estimated emissions reduction: 1,090 metric tons CO2e annually.\n- Can sustain critical refrigeration for approximately 10 hours without grid power under typical conditions.\n- Fixed-price EPC contract; completion in 11 months.\n- Vendor has completed nine comparable cold-storage projects.\n- Roof reinforcement is included, but final structural verification remains a condition before notice to proceed.\n\nPROPOSAL B — PINE STATE ENERGY\n- 1.0 MW solar, 2.4 MWh battery, and 1.1 MW renewable-diesel generator.\n- Gross capital cost: $6.35 million.\n- Confirmed federal grant: $1.25 million.\n- Annual Year 1 electricity savings: $495,000.\n- Annual Year 1 O&M: $128,000.\n- Battery replacement: $720,000 at end of Year 10.\n- Mitigates 96% of modeled long-outage exposure, assuming fuel delivery remains available.\n- Estimated emissions reduction: 820 metric tons CO2e annually under expected generator use.\n- Can sustain critical refrigeration for at least 48 hours with a full fuel tank.\n- Completion in 14 months.\n- Vendor has completed three comparable cold-storage projects and 22 hospital microgrids.\n- Contract price may increase by up to $300,000 if interconnection protection upgrades are required; assign a 40% probability to this cost and include its expected value in the base-case economics.\n\nPROPOSAL C — ATLANTIC STORAGE SYSTEMS\n- 4.8 MWh battery with no on-site generation.\n- Gross capital cost: $3.45 million.\n- No grant.\n- Annual Year 1 electricity savings: $430,000.\n- Annual Year 1 O&M: $52,000.\n- Battery replacement: $1.30 million at end of Year 9.\n- Mitigates 52% of modeled long-outage exposure.\n- Estimated emissions reduction: 210 metric tons CO2e annually.\n- Can sustain critical refrigeration for approximately eight hours at current loads.\n- Completion in seven months.\n- Vendor has completed 31 commercial battery projects, but none at refrigerated warehouses.\n- The proposal guarantees 90% usable battery capacity through Year 8; there is no performance guarantee after replacement.\n\nDEFERRAL OPTION\n- No Year 0 capital outlay.\n- The procurement adviser expects equipment and labor pricing to rise 6% over the next year.\n- Tidal Gridworks’ state grant is reserved for nine months and would probably be lost if the decision is deferred.\n- Pine State Energy’s federal grant remains available for 18 months.\n- No vendor will hold its current completion schedule.\n- Outage exposure and current electricity costs continue during the delay.\n\nREQUIRED DELIVERABLE\nWrite a polished memo of approximately 1,500–1,900 words with these sections:\n1. To / From / Date / Subject header.\n2. Executive recommendation stating the decision in the first paragraph.\n3. Decision rationale, explaining why the recommended option best fits the board’s ordered priorities.\n4. A compact comparison table covering net upfront cost, Year 1 net operating benefit, annualized avoided outage loss, base-case NPV, simple payback, approximate IRR, outage capability, emissions reduction, schedule, and major execution risk.\n5. Financial analysis describing the cash-flow method and separating electricity savings, O&M, avoided expected inventory loss, replacement costs, and proposal-specific expected costs. Define simple payback as the first year cumulative undiscounted net cash flow becomes positive.\n6. Sensitivity analysis for the recommended proposal under three separate downside cases: electricity savings are 20% below forecast; no avoided-loss benefit is credited; and upfront cost is 10% higher. Report the revised NPV for each case and explain what each reveals.\n7. Key risks and mitigations, with named owners by role, such as COO, CFO, facilities director, or general counsel.\n8. A 90-day action plan with concrete milestones.\n9. A final board resolution written as one sentence that can be copied into meeting minutes.\n\nMaintain a decisive, sober tone. Explain technical concepts in plain English without oversimplifying the economics. Distinguish calculated facts from judgment. Do not recommend deferral merely because information is incomplete; assess whether the identified uncertainties can be managed through contract terms, diligence, contingencies, or approval conditions."
    }
  ]
}