Make a tamper-evident meeting record

# Assemble the Decosa tamper-evident record on this machine

You are setting up a self-hosted tamper-evident record on this Linux machine: live captions, a list of motions, votes and actions, then (when a session stops) a speaker-attributed transcript, a summary where every sentence cites transcript lines, a claim check, and a session record signed with this machine's own key. Work step by step, show me each command before running anything that needs sudo, and stop if a check fails.

**Before anything else, remind me:** tell everyone that a session is being recorded and get consent where the law requires it. This is not a certified court record. Repeat both points in your final summary.

## Step 0: set up with a coding agent, rehearse on mock data, then go private

This prompt is for a coding agent running on the machine that will host the service. We recommend Claude Code with
Claude Opus 5.5; any capable coding agent works. Work in this order:

1. Set up on mock data only. During the whole setup you (the agent) work with the synthetic sample bundle below and
   nothing else. Do not ask me for real data, and do not open, read, list or copy files that hold real data, even to
   "test with something realistic".
2. Rehearse. When the steps below are done and the service is healthy, fetch the mock-data bundle for this tool,
   https://decosa.ai/samples/record.zip (630 KB, 9 checks, synthetic or openly licensed: see `licence` in expected.json),
   show me what is in it, and run the rehearsal against the local API:
   `docker compose exec api python scripts/rehearse.py record` (the api image carries the same bundle under /app/rehearsal/record/;
   with no key set, the script asks the local API for a short demo token). From a decosa-api checkout instead:
   `python scripts/rehearse.py record --bundle record.zip --base-url http://127.0.0.1:<PORT>`.
   It sends the mock inputs to the local API and prints PASS or FAIL for each expected property (for example: "the session ends with a done event", "the session reports no errors", "final captions arrive"). Show me
   the full output. Every check must pass. If one fails, fix the install and run it again; never edit `expected.json`
   to make a check pass.
3. Stop there. Once the rehearsal passes, tell me, and I will run my own data against the local API myself, on this
   machine.

For the person running this: a coding agent that runs in the cloud sees everything in its context, including files it
reads, command output and anything pasted into the chat. Keep real data out of the chat and out of anything the agent
can read. Switch to your own data only after the rehearsal has passed and the agent's work is done.

## What you are building

| service | image | model | port |
|---|---|---|---|
| `llm` | `${DECOSA_REGISTRY}/decosa-llm:0.1.0` (vLLM 0.29.0, `vllm/vllm-openai@sha256:c2914767605584b6d8f45686b82de173ecc99e781897aa3d0a66dacd72c51ae1`) | `nvidia/Qwen3.8-27B-NVFP4` @ `482ca0f3832238542f8f5295dde86b5f22711d80`, Apache-2.0 | internal 8000 |
| `asr` | `${DECOSA_REGISTRY}/decosa-asr:0.1.0` (vLLM 0.27.1 + `mistral-common[audio]`) | `mistralai/Voxtral-Mini-4B-Realtime-2602`, Apache-2.0 | internal 8000 |
| `api` | `${DECOSA_REGISTRY}/decosa-api:0.1.0` (no GPU) | none | `127.0.0.1:8445` |
| `diarize` (optional) | built from the decosa-api source (`services/diarize`) | `OpenMOSS-Team/MOSS-Transcribe-Diarize` @ `704aa4a9c304e8520be88901e0d1960158ef5b15`, Apache-2.0 | internal 8092 |

Audio, captions, transcripts and records stay on this machine. Without `diarize`, records still work: the live captions become the transcript lines, without speaker labels.

## 1. Check the GPU, driver and Docker

1. Run `nvidia-smi`. I need one NVIDIA GPU with at least 48 GB and driver 580 or newer.
   - Blackwell (RTX PRO 6000, B200): use the defaults below (NVFP4).
   - Hopper (H100/H200): set `LLM_MODEL=Qwen/Qwen3.8-27B-FP8`, `LLM_REVISION=main`, `LLM_GPU_UTIL=0.62`.
   - 48 GB Ada/L40S: FP8 checkpoint as above, plus `LLM_MAX_LEN=16384`, `LLM_GPU_UTIL=0.70`, `ASR_GPU_UTIL=0.22`, `DECOSA_LIVE_CAP=2`.
   - Under 48 GB: stop and tell me it will not fit.
   Only the Blackwell defaults have been measured; the other rows are starting points.
2. Check `docker --version` and `docker compose version`. If Docker is missing, install Docker Engine from Docker's official apt/dnf repository for this distro.
3. Check `docker run --rm --gpus all ubuntu nvidia-smi`. If it fails, install the NVIDIA Container Toolkit (`nvidia-container-toolkit`) from NVIDIA's repository, run `sudo nvidia-ctk runtime configure --runtime=docker` and restart Docker.
4. Confirm about 80 GB of free disk for images and weights.

## 2. Get the images

The images are **on request** while self-host is in early access: ask at https://decosa.ai/contact?topic=self-host, and Decosa sends the registry (set it as `DECOSA_REGISTRY`), pull access and the compose file. Try `docker pull ${DECOSA_REGISTRY}/decosa-{llm,asr,api}:0.1.0`. If a pull fails (not published yet, or no access), build from source once the `decosa-api` source is published: clone it, then `docker compose build llm asr api` in the repo, which builds the same tags from `docker/`. If neither works, stop and tell me.

## 3. Write the compose file

Create `~/decosa-record/.env`:

```bash
DECOSA_TAG=0.1.0
DECOSA_GPU=0
LLM_MODEL=nvidia/Qwen3.8-27B-NVFP4
LLM_REVISION=482ca0f3832238542f8f5295dde86b5f22711d80
LLM_MAX_LEN=65536
LLM_GPU_UTIL=0.60
ASR_GPU_UTIL=0.25
DECOSA_LLM_ROUTE=direct   # local model; receipts are signed by this box's own key ("attested")
DECOSA_SIGNER_NAME="<who signs these records, e.g. Riverbend town clerk>"
DECOSA_LIVE_CAP=4
```

Create `~/decosa-record/docker-compose.yml` with exactly these services:

```yaml
name: decosa-record
x-gpu: &gpu
  deploy: { resources: { reservations: { devices: [ { driver: nvidia, device_ids: ["${DECOSA_GPU:-0}"], capabilities: [gpu] } ] } } }
x-health: &health
  interval: 15s
  timeout: 5s
  retries: 5
services:
  llm:
    image: ${DECOSA_REGISTRY}/decosa-llm:${DECOSA_TAG}
    <<: *gpu
    ipc: host
    restart: unless-stopped
    volumes: [hf-cache:/root/.cache/huggingface]
    command: ["${LLM_MODEL}", "--revision", "${LLM_REVISION}", "--served-model-name", "qwen3.8-27b",
              "--language-model-only", "--max-model-len", "${LLM_MAX_LEN}", "--gpu-memory-utilization", "${LLM_GPU_UTIL}",
              "--max-num-seqs", "16", "--kv-cache-dtype", "fp8_e4m3", "--speculative-config", '{"method":"mtp","num_speculative_tokens":3}',
              "--seed", "0", "--enable-force-include-usage", "--disable-uvicorn-access-log", "--host", "0.0.0.0", "--port", "8000"]
    healthcheck: { <<: *health, test: ["CMD", "python3", "-c", "import urllib.request; urllib.request.urlopen('http://127.0.0.1:8000/health', timeout=4)"], start_period: 900s }
  asr:
    image: ${DECOSA_REGISTRY}/decosa-asr:${DECOSA_TAG}
    <<: *gpu
    ipc: host
    restart: unless-stopped
    depends_on: { llm: { condition: service_healthy } }   # start after llm so the memory split is stable
    volumes: [hf-cache:/root/.cache/huggingface]
    command: ["--model", "mistralai/Voxtral-Mini-4B-Realtime-2602", "--tokenizer-mode", "mistral", "--config-format", "mistral",
              "--load-format", "mistral", "--compilation-config", '{"cudagraph_mode":"PIECEWISE"}', "--max-model-len", "45000",
              "--max-num-batched-tokens", "8192", "--max-num-seqs", "16", "--gpu-memory-utilization", "${ASR_GPU_UTIL}",
              "--served-model-name", "voxtral-realtime", "--host", "0.0.0.0", "--port", "8000"]
    healthcheck: { <<: *health, test: ["CMD", "python3", "-c", "import urllib.request; urllib.request.urlopen('http://127.0.0.1:8000/health', timeout=4)"], start_period: 600s }
  api:
    image: ${DECOSA_REGISTRY}/decosa-api:${DECOSA_TAG}
    restart: unless-stopped
    depends_on: { llm: { condition: service_healthy }, asr: { condition: service_healthy } }
    environment:
      DECOSA_ASR_WS: ws://asr:8000/v1/realtime
      DECOSA_LLM_ROUTE: ${DECOSA_LLM_ROUTE}
      DECOSA_LLM_URL: http://llm:8000/v1
      DECOSA_LLM_MODEL: qwen3.8-27b
      DECOSA_LIVE_CAP: ${DECOSA_LIVE_CAP}
      DECOSA_SESSIONS_PER_IP_HOUR: "1000"
      DECOSA_BUDGET_AUDIO_S: "3600"
      DECOSA_BUDGET_LLM_TOKENS: "200000"
      DECOSA_SESSION_TTL_S: "28800"
      DECOSA_LOCAL_SIGNING: "on"                      # Ed25519 key created at /data/attest/ed25519.pem on first start
      DECOSA_SIGNER_NAME: ${DECOSA_SIGNER_NAME}
      DECOSA_DIARIZE_URL: ""                          # step 4 sets this if you add speaker labels
      DECOSA_CORS_ORIGIN_REGEX: '^https?://(localhost|127\.0\.0\.1)(:\d+)?$$'
    ports: ["127.0.0.1:8445:8445"]
    volumes: [decosa-data:/data]
    healthcheck: { <<: *health, test: ["CMD", "python", "-c", "import urllib.request; urllib.request.urlopen('http://127.0.0.1:8445/healthz', timeout=4)"], start_period: 20s }
volumes: { hf-cache: {}, decosa-data: {} }
```

Run `docker compose up -d`, then poll `docker compose ps` until all three are healthy (the LLM takes 5–10 minutes the first time) and `curl -s localhost:8445/healthz` shows `"asr": true, "llm": true`. If `llm` runs out of memory, lower `LLM_GPU_UTIL` or `LLM_MAX_LEN`; the two GPU shares must add up to less than about 0.9.

## 4. Optional: speaker labels (diarize service)

The `diarize` service (MOSS-Transcribe-Diarize 0.9B, Apache-2.0) has no published image yet. If I want speaker labels, run `services/diarize` from the decosa-api source on this box (install with the `uv` commands at the top of `services/diarize/requirements.txt`, then `DIARIZE_HOST=172.17.0.1 DIARIZE_PORT=8092 DIARIZE_DEVICE=cuda:0 .venv-diarize/bin/python services/diarize/server.py`; `172.17.0.1` is the docker bridge address from `ip -4 addr show docker0`, reachable from containers and not from the LAN). Add `extra_hosts: ["host.docker.internal:host-gateway"]` to `api` and set `DECOSA_DIARIZE_URL: http://host.docker.internal:8092`. `curl -s localhost:8445/healthz` then shows `"diarize": true`. The 0.9B model needs about 2 GB of weights; lower `LLM_GPU_UTIL` to 0.55 so it fits. That fit is an estimate, not measured. Without it the record still works: `done.summary.pass2` reads `unavailable` and the transcript lines have no speaker names. If the source isn't available, skip this step.

## 5. The signing key

1. `curl -s localhost:8445/attest/signing-key` returns `{"scheme":"ed25519","pubkey":"<64 hex>","key_id":...,"name":...}`. Show me the `pubkey`.
2. Tell me to **back up the `decosa-data` volume**: it holds the private key (`/data/attest/ed25519.pem`, mode 0600). With a new key, new records still verify, but old records no longer show as issued by this box.
3. Tell me to publish the public key where people will check my records (website, agenda, case file).
4. Receipts from this box say `status: "attested"`: signed by our own key. They prove that nothing was changed after signing and who signed. They do not prove that the model ran or heard correctly.

## 6. Smoke test: record, verify, tamper

```bash
TOKEN=$(curl -s localhost:8445/demo/session -H 'content-type: application/json' -d '{"vertical":"record"}' | jq -r .token)
curl -sN localhost:8445/demo/replay -H "authorization: Bearer $TOKEN" -H 'content-type: application/json' \
  -d '{"vertical":"record","script_id":"record-council-meeting"}' > /tmp/record.sse
grep -oE '"(type|lane|kind)": "[a-z_]+"' /tmp/record.sse | sort | uniq -c
grep '"type": "done"' /tmp/record.sse | sed 's/^data: //' | jq '.summary.record' > /tmp/record.json
curl -s localhost:8445/record/verify -H 'content-type: application/json' --data-binary @/tmp/record.json | jq '{ok, summary}'
jq '(.entries[] | select(.kind=="line") | .text) |= sub("the"; "not the")' /tmp/record.json > /tmp/edited.json
curl -s localhost:8445/record/verify -H 'content-type: application/json' --data-binary @/tmp/edited.json | jq '{ok, first_bad, summary}'
```

The meeting is 91.7 s of synthetic audio, paced in real time. Pass if:
- the stream has `ready` with `asr_receipts: "attested"`, `transcript` events, `receipt` events with `kind: "asr"` and with `status: "attested"`, and lanes `actions`, `chain`, `final_transcript`, `summary`, `verifier`, `record`;
- the first verify prints `ok: true`;
- the second prints `ok: false`, with `first_bad` at the first edited transcript line.

The other script is `record-investigation-interview` (58.8 s).

## 7. Point the app at the local API

- Base URL: `http://localhost:8445` (for the web app, `NEXT_PUBLIC_DECOSA_API=http://localhost:8445`). Add the app's origin to `DECOSA_CORS_ORIGINS` if it is not localhost.
- `POST /demo/session {"vertical":"record"}` returns a token. Open `ws://localhost:8445/ws/live?vertical=record&token=<t>`, send 16 kHz mono PCM16 little-endian frames of about 100 ms, then `{"type":"stop"}`. Save `done.summary.record` as JSON: that file is the record. Close code `4403` means the origin is not allowed.
- Anyone can check a record: `POST /record/verify`, or the "Verify a record" panel on the Decosa site, which checks it in the browser.
- Keep the API on `127.0.0.1`. To reach it from the LAN, put a TLS reverse proxy in front and set `DECOSA_TRUSTED_PROXIES`.

Finish with a summary: what is running, the health output, the signing key's pubkey, both verify results, and the consent and not-a-court-record reminders.