Read a demand before the deadline · Insurance

Use it from your codeThe hosted API with your key, and prompts to paste into a coding agent

Self-host · your GPUs · recommended

Run it yourself, on request

The same open models and app, on 1x RTX PRO 6000 (96 GB) for Qwen3.8-27B; the deadline rules and the tie-out need no GPU.
Data never leaves your machines, and there are no Decosa charges.
One prompt for Claude Code or Codex assembles the whole stack.
Early access: the container images are not public yet and the source needs access; the prompt says how to ask.

Get the assembly prompt

Hosted · by Decosa

Get an API key

Call the injury demand reader API from your own code in minutes.
Every model answer carries a signed receipt.
Synthetic, public or test data only: real confidential data belongs on your own hardware.

Build with it

Paste one of these into Claude Code, Codex or another coding agent. The first wires your project to the hosted API with your DECOSA_API_KEY. The second pulls our containers and runs the same stack on your own GPU, with no Decosa charges.

Base URL: https://api.decosa.ai
Auth: Authorization: Bearer $DECOSA_API_KEY (or a demo session token)
Tool id: demand-reader

Use the hosted API

# The Decosa demand reader: use the hosted API

You are wiring the Decosa demand reader into this project. It returns:
- the respond-by date the demand letter sets, computed in code from its own words and the date you received it;
- the statute's minimum for the encoded states (CA, GA, MO, MT, UT; Florida's safe harbour), each with its source;
- every condition of acceptance, quoted with its page, and the state's required elements;
- each bill line tied to its page, summed, and reconciled with what the letter claims;
- providers billed with no records, records with no bill, gaps in care and earlier care;
- notes on the letter's citations (a statute from another state, a case that isn't at its cite).

Every model call has a signed receipt. Use only what is listed below; if you need something else, stop and ask me.

- Base URL: `https://api.decosa.ai`
- Health check: `GET https://api.decosa.ai/healthz`.
- **Hosted use is for made-up files only.** Real files belong on a self-hosted box (see the self-host prompt) or confidential access. Say so wherever this is wired in.
- It never values a claim or recommends paying, offering or accepting. Deadlines are reminders from the letter's own words; check them against the file and, where it matters, with counsel. States other than CA, GA, MO, MT, UT and FL return "not encoded: check with counsel".

## Auth: API key (or a demo session)
1. Preferred: an API key (`dk_…`) from "Get an API key" on the tool page. Keep it in `DECOSA_API_KEY`, never in code. Send `Authorization: Bearer $DECOSA_API_KEY`.
2. Without a key: `POST https://api.decosa.ai/demo/session` with `{"vertical": "demand-reader"}` returns `{"token", "expires_at", "budget"}`. Demo sessions are limited per IP per hour; over a limit you get HTTP 429 with `Retry-After`. A token runs one request at a time (409 otherwise).

## Endpoints
- `POST /demand/read` (token). Body: `{state, received: {date, method}, claim?: {claim_number?, loss_date?}, documents: [{kind: letter|bill|record, title?, provider?, pages: [text, ...]}]}`, or `{"sample_id": "..."}` for a sample from `GET /demand/samples`. Add `"stream": true` (or `Accept: text/event-stream`) for events; the last one is `result`, which carries `usage` (tokens and cost at list price).
- `GET /demand/info`: what is checked, sources, limits, not checked.
- `POST /record/verify` (no auth): re-check the signed record.

Show the result's own words to the user: the verdict first, then each item with its quote and where it is. Never add a value, a recommendation or a "safe" of your own.

Run it yourself (containers)

On request. The container images and the compose file aren’t public yet. Ask for self-host access and Decosa sends the registry (DECOSA_REGISTRY) and the compose file’s URL (DECOSA_COMPOSE_URL) these steps use. They are the steps we tested end to end on a fresh machine.

# The Decosa demand reader: run it yourself (containers)

You are setting up the Decosa demand reader on this machine, so the files never leave it. It returns:
- the respond-by date the demand letter sets, computed in code from its own words and the date you received it;
- the statute's minimum for the encoded states (CA, GA, MO, MT, UT; Florida's safe harbour), each with its source;
- every condition of acceptance, quoted with its page, and the state's required elements;
- each bill line tied to its page, summed, and reconciled with what the letter claims;
- providers billed with no records, records with no bill, gaps in care and earlier care;
- notes on the letter's citations (a statute from another state, a case that isn't at its cite).

Nothing is sent to Decosa's hosted API.

Status: the container images (${DECOSA_REGISTRY}/decosa-*) and the compose file are on request while self-host is in early access (not on a public registry yet): ask at https://decosa.ai/contact?topic=self-host, and Decosa sends the registry as DECOSA_REGISTRY, the compose file URL as DECOSA_COMPOSE_URL, and pull access. If a pull fails with "not found", "denied" or "unauthorized", stop and tell me. Do not substitute other images.

Ask me before any command that needs sudo, and show me the command first.

## 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/demand-reader.zip (4 KB, 8 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 demand-reader` (the api image carries the same bundle under /app/rehearsal/demand-reader/;
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 demand-reader --bundle demand-reader.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: "it is a time-limited demand", "respond by 3 Sep 2026, from the letter's 30 days after receipt", "the statute's minimum is the same day"). 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.

## Steps
1. Docker: if `docker compose version` fails, install Docker Engine and the compose plugin (docs.docker.com/engine/install), and the NVIDIA container toolkit; check `docker run --rm --gpus all nvidia/cuda:12.8.0-base-ubuntu24.04 nvidia-smi`.
2. Fetch the compose file: `mkdir -p ~/decosa && cd ~/decosa && curl -fsSL "${DECOSA_COMPOSE_URL}" -o compose.yaml`. Keep the `llm` service (Qwen3.8-27B on vLLM) and the `api` service; on `api` set `DECOSA_LLM_ROUTE=direct`, `DECOSA_LLM_URL=http://llm:8000/v1`, `DECOSA_LLM_MODEL=qwen3.8-27b`; keep its data on a named volume; bind every port to 127.0.0.1.
3. Pull and start: `docker compose pull && docker compose up -d`. Wait for the `llm` health check (the first start downloads about 20 GB of weights).
4. Check: `curl -fsS http://127.0.0.1:<PORT>/demand/info` answers; `GET /attest/signing-key` shows this box's public key.
5. Smoke test: get a token with `POST /demo/session {"vertical":"demand-reader"}`, then `POST /demand/read {"sample_id": "ga-policy-limits"}`. Expect:
- `.deadline.respond_by` is `"2026-09-03"` and `.deadline.statute_minimum.date` is `"2026-09-03"` (30 days from receipt, O.C.G.A. 9-11-67.1);
- an item says `$19,540.00` of the `$31,450.00` claimed is not supported by any bill;
- a note says the letter gives 10 days where the statute gives at least 30;
- no item or the memo values the claim;
- every receipt has `"status": "attested"` and the record verifies.

Finish with a summary and these reminders:
- Demand packages hold medical records and personal information. Keep everything on this machine: the model route stays local (`direct`).
- It never values a claim or recommends paying, offering or accepting. Deadlines are reminders from the letter's own words; check them against the file and, where it matters, with counsel. States other than CA, GA, MO, MT, UT and FL return "not encoded: check with counsel".

Run it on your own hardwareWhat it needs, and the prompt that sets it up

Run it on your own GPU

Same app, same pinned models, your hardware. Nothing goes to our servers and there are no Decosa charges.

Hardware check

Check your own hardware

CPU only, 64 GB RAMDoesn't fit
Qwen3.8-27B (NVIDIA NVFP4) needs a GPU.
GeForce RTX 4090standard tierRuns
The standard tier fits with changes: Replace Qwen3.8-27B (NVIDIA NVFP4) with A community 4-bit build of Qwen3.8-27B (AWQ or GGUF). This build is NVIDIA NVFP4, which needs a Blackwell GPU. (Memory is an estimate.)
GeForce RTX 5090standard tierRuns
The standard tier fits with changes: Qwen3.8-27B (NVIDIA NVFP4): run it at its smallest setting (about 28 GB instead of 57.6 GB), with a shorter context and fewer parallel sessions.
2x GeForce RTX 5090standard tierRuns
The standard tier fits with changes: Split the language model across the GPUs with tensor parallelism (vLLM --tensor-parallel-size).
L40Sstandard tierRuns
The standard tier fits with changes: Replace Qwen3.8-27B (NVIDIA NVFP4) with Qwen3.8-27B official FP8. This build is NVIDIA NVFP4, which needs a Blackwell GPU.
H100 80 GB (SXM)standard tierRuns
The standard tier fits with changes: Replace Qwen3.8-27B (NVIDIA NVFP4) with Qwen3.8-27B official FP8. This build is NVIDIA NVFP4, which needs a Blackwell GPU.
RTX PRO 6000 Blackwell 96 GBstandard tierRuns
The standard tier fits (57.6 of 96 GB).
2x RTX PRO 6000 Blackwell 96 GBbest tierRuns
The standard tier fits (57.6 of 192 GB). The best tier fits too.
Apple M3 Ultra (Mac Studio), 96 GBstandard tierRuns
The standard tier fits with changes: Replace Qwen3.8-27B (NVIDIA NVFP4) with Qwen3.8-27B MLX 4-bit. MLX build for Apple Silicon.
Apple M5 Max, 64 GBstandard tierRuns
The standard tier fits with changes: Replace Qwen3.8-27B (NVIDIA NVFP4) with Qwen3.8-27B MLX 4-bit. MLX build for Apple Silicon.

Memory per component comes from measured footprints, the tool's stack.json, or an estimate from its parameter count, and each is labelled that way below. Only an RTX PRO 6000 and an M3 Ultra Mac Studio have actually been run.

On request. The container images and the compose file aren’t public yet. Ask for self-host access and Decosa sends the registry (DECOSA_REGISTRY) and the compose file’s URL (DECOSA_COMPOSE_URL) these steps use. They are the steps we tested end to end on a fresh machine.

1
Check the GPU, Docker and the NVIDIA Container Toolkit
The driver must see the GPU, and Docker must be able to pass it into a container.
```
nvidia-smi
docker compose version
docker run --rm --gpus all ubuntu nvidia-smi
```
2
Fetch the compose file
One file describes the API and the language model as services.
```
mkdir -p ~/decosa && cd ~/decosa
curl -fsSL "${DECOSA_COMPOSE_URL}" -o compose.yaml
```
3
Pull and start
The first start downloads pinned model weights, tens of gigabytes.
```
docker compose pull
docker compose up -d
```

4

Check health

Wait until the API reports ok with the language model loaded. Then point your app at the local base URL.

curl -fsS http://localhost:<PORT>/healthz
# {"ok": true, "llm": true, ...}
curl -fsS -X POST http://localhost:<PORT>/demo/session \
  -H 'Content-Type: application/json' -d '{"vertical":"demand-reader"}'

Set up with a coding agent, rehearse on mock data, then go private

Set up with a coding agent. Paste the self-host prompt into a coding agent on the machine that will run the service. We recommend Claude Code with Claude Opus 5.5; any capable coding agent works.
Rehearse on mock data. The agent runs the tool on a bundle of synthetic inputs and checks each answer against the bundle's expected.json. Every check must print PASS.
Go private. Only then do you run your own data against the local API, yourself, on that machine. Never give the agent real data during setup: a coding agent that runs in the cloud sees everything in its context, so keep real data out of the chat and out of the files it reads.

Rehearsal command

docker compose exec api python scripts/rehearse.py demand-reader

Download the mock-data bundle (4 KB, 8 checks)expected.json

A synthetic Georgia motor-vehicle demand (two letter pages), two bills and three sets of records, received by certified mail on 4 Aug 2026. The respond-by date must be 3 Sep 2026 (30 days from receipt, O.C.G.A. 9-11-67.1), the $19,540.00 of the $31,450.00 claimed that no bill supports must be found, the letter's '10 days' misstatement and its Florida statute must be noted, and the signed record must verify.

What the rehearsal checks

it is a time-limited demand
respond by 3 Sep 2026, from the letter's 30 days after receipt
the statute's minimum is the same day
the bills total $11,910.00
$19,540.00 claimed is not supported by any bill
the signed record verifies
a record with its respond-by date changed no longer verifies
every model call has a signed receipt

Licence: Synthetic: the claimant, insured, firm, insurer and providers are invented. Part of decosa-api.

Prompt for your coding agent