Last updated: 2026-09-04 Tracks known gaps between the live API and published documentation.
vLLM chat + vLLM embeddings + srt-data-layer
└── srt-artemis (api.solidrust.ai)
└── docs.solidrust.ai (this repo)
LiteLLM is not in this path. It was expelled; do not put it back in diagrams.
The docs site documents the public surface of api.solidrust.ai.
Internal-only endpoints (admin, connectors, tasks, hygiene, etc.) are
intentionally not documented.
| Component | Status | Notes |
|---|---|---|
| Chat model copy | ✅ 2026-09-04 | vllm-primary / Gemma 4 12B QAT / 16k |
| Embeddings copy | ✅ 2026-09-04 | Qwen/Qwen3-Embedding-0.6B / 1024-dim |
| Rate limits | ✅ 2026-09-04 | Free/Pro/Enterprise from PAM tier-limits.ts |
| Failover copy | ✅ 2026-09-04 | No LiteLLM; agent-only OpenRouter |
| Internal admin paths | Intentionally undocumented | 97 live paths vs curated public spec |
| Endpoint Group | Purpose | Public Doc Priority |
|---|---|---|
/data/v1/connectors/* |
Data connector management | Low |
/data/v1/tasks/* |
Celery task queue management | Low |
/data/v1/analytics/* |
Query analytics dashboard | Low |
/data/v1/memory/* |
Memory agent | Medium |
/data/admin/* |
Dead-letter queue | None |
- First-party Python SDK — #1
- First-party JS SDK — #2
- Video tutorials — #3 (no footage; page is a stub)
- Customer ingest tenancy — #4 (owner: srt-data-layer)
- Clarity / SearXNG / Firecrawl public API — #5
-
sources.md/stats.mdreference pages - Document memory endpoints if they become public
- CI: fail if docs still mention
bge-m3,qwen3-4b, or LiteLLM as live