API

Demo chat.

A public, unauthenticated endpoint backed by a fixed engagement context. Useful for prototypes and marketing integrations. Do not rely on it for production traffic.

Contexts never mix

The demo endpoint uses a hardcoded DEMO_SYSTEM_PROMPT that describes a fictional manufacturing engagement (Vantage Industrial Group). It never reads from your organization's database rows, and the live endpoint never falls through to the demo prompt. Do not proxy one to the other.

Contract

POST/api/chat/demoStable

Stream an assistant reply conditioned on the fixed Vantage Industrial Group engagement.

AuthNone

Rate15 req / hour per IP

Request body

messagesrequired

MessageParam[]

Same shape as live chat — non-empty array; final message must have role "user". The server does not reconstruct history, so clients should include prior turns in the array for conversational continuity.

Example request

curl -N -X POST https://app.corvanahq.com/api/chat/demo \
  -H "Content-Type: application/json" \
  -d '{
    "messages": [
      { "role": "user", "content": "Walk me through Vantage in two sentences." }
    ]
  }'

Example response

200 OK — streamed plain text

text

Margaret, here's the short version — we've surfaced roughly $14.2M in …

Response is plain text chunks, same shape as live chat.

Error codes

Code	Status	Description
`INVALID_JSON`	400	Request body could not be parsed.
`MISSING_MESSAGES`	400	`messages` missing or empty.
`RATE_LIMITED`	429	15 requests per hour per client IP — see Retry-After.
`UPSTREAM_OVERLOADED`	503	Anthropic overloaded.
`UPSTREAM_ERROR`	502	Other model-side error.

Differences from live chat

No authentication. The demo endpoint does not check session cookies.
No persistence. No Conversation or ChatMessage rows are written.
No conversation header. X-Conversation-Id is not set.
Lower budget. max_tokens is 2048 (live is 4096) and output effort is medium (live is high). Answers are shorter by design.
Fixed prompt. The system prompt is a single cached text block describing Vantage. L4–L6 dynamic layers are not composed.
IP rate limit. 15 requests per hour per source IP, derived from the first token of x-forwarded-for (or x-real-ip). See Rate limits.

When to use the demo endpoint

Prototyping a UI integration before provisioning an organization.
Marketing or sales surfaces that need a no-sign-in conversational demo.
Public benchmarking of streaming behavior, time-to-first-byte, or formatting fidelity.

For anything that needs an organization's actual context, engagement memory, or compliance posture — use the live endpoint.

Previous: Live chat · Next: Errors.

Contract

POST/api/chat/demoStable

Stream an assistant reply conditioned on the fixed Vantage Industrial Group engagement.

AuthNone

Rate15 req / hour per IP

Request body

messagesrequired

MessageParam[]

Example request

curl -N -X POST https://app.corvanahq.com/api/chat/demo \
  -H "Content-Type: application/json" \
  -d '{
    "messages": [
      { "role": "user", "content": "Walk me through Vantage in two sentences." }
    ]
  }'

Example response

200 OK — streamed plain text

text

Margaret, here's the short version — we've surfaced roughly $14.2M in …

Response is plain text chunks, same shape as live chat.

Error codes

Code	Status	Description
`INVALID_JSON`	400	Request body could not be parsed.
`MISSING_MESSAGES`	400	`messages` missing or empty.
`RATE_LIMITED`	429	15 requests per hour per client IP — see Retry-After.
`UPSTREAM_OVERLOADED`	503	Anthropic overloaded.
`UPSTREAM_ERROR`	502	Other model-side error.

Differences from live chat

No authentication. The demo endpoint does not check session cookies.

No persistence. No Conversation or ChatMessage rows are written.

No conversation header. X-Conversation-Id is not set.

Lower budget. max_tokens is 2048 (live is 4096) and output effort is medium (live is high). Answers are shorter by design.

Fixed prompt. The system prompt is a single cached text block describing Vantage. L4–L6 dynamic layers are not composed.

IP rate limit. 15 requests per hour per source IP, derived from the first token of x-forwarded-for (or x-real-ip). See Rate limits.

When to use the demo endpoint

Prototyping a UI integration before provisioning an organization.

Marketing or sales surfaces that need a no-sign-in conversational demo.

Public benchmarking of streaming behavior, time-to-first-byte, or formatting fidelity.

For anything that needs an organization's actual context, engagement memory, or compliance posture — use the live endpoint.