Ragnarock AI
Use the unified AI assistant for requirements, planning, and engineering help.
Ragnarock (/dashboard/projects/{projectId}/ragnarock) is the unified chat surface. It uses dedicated REST endpoints and can stream updates over WebSocket (/ai-chat namespace).
Agent capabilities
| Agent key | What it does | Typical prompt |
|---|---|---|
requirements | Clarifies and structures an SRS | "Turn this brief into user stories with acceptance criteria" |
developer_intelligence | Answers using SRS + linked repos | "How should we structure the auth module?" |
project_planner | One-shot task breakdown | Triggered via plan generation API |
qa_intelligence | Test scenarios and suites | "List edge cases for the login flow" |
change_impact | Change analysis (seeded; routing TBD) | "What breaks if we remove this API?" |
Agents are registered in Postgres (AgentDefinition) and executed by the FastAPI worker via Redis Streams.
User workflow
A session is created via POST /projects/:projectId/ai/ragnarock/sessions (or resumed from the
list).
Send a message — POST .../ai/ragnarock/chat returns 202 Accepted while the worker
processes the job.
Listen for completion via polling or the /ai-chat WebSocket gateway (requirements sessions
use the same broadcast pattern).
Related API endpoints
Base path: /projects/:projectId/ai
| Method | Path | Purpose |
|---|---|---|
POST | /ragnarock/sessions | Create unified chat session |
GET | /ragnarock/sessions | List sessions |
GET | /ragnarock/sessions/:sessionId | Messages for session |
POST | /ragnarock/chat | Send message (async) |
POST | /chat/sessions | Requirements-specific session |
POST | /requirements | Text/URL requirements turn |
POST | /requirements/upload | PDF/DOCX upload (max 15 MB) |
POST | /arch-doc/generate | Architecture document generation |
POST | /plan/generate | Planner one-shot |
POST | /qa/generate | QA test suite generation |
GET | /draft | Project shared SRS draft |
Async processing
Streams (defaults from .env.example):
- Jobs:
stream:ai-requirements:jobs - Results:
stream:ai-requirements:results
Tips for better answers
- Set the correct persona on your project membership.
- Link Repositories before asking implementation questions.
- Maintain project Skills with coding standards and glossary terms.
- Keep the project draft SRS current — new requirement chats continue from it (
GET /ai/draft).
Developer details: AI agents. Troubleshooting: Common issues.