OpenRouter multi-model adapter and Voyage AI semantic search added in one PR

Chris-o-O adds five additional LLMs through a single OpenRouter key and bolts on pgvector-backed semantic search using Voyage's legal-domain embedding model. Both are the two most-requested gaps in the upstream codebase, landed together in commit `3279a584`.

searchintegration

The OpenRouter adapter (backend/src/lib/llm/openrouter.ts) plugs into the existing Claude/Gemini provider abstraction - same pattern, different endpoint. GPT-4o, GPT-4 Turbo, Mistral Large, Llama 3.3 70B, and Gemma 3 27B are added to the model roster via models.ts. A per-user openrouter_api_key column lands in user_profiles; the frontend wires it through ModelToggle and account settings.

Semantic search uses Voyage AI's voyage-law-2 (1024-dim, legal-tuned) rather than the default OpenAI or Cohere embedder. Documents are chunked at fixed 800-token windows with 150-token overlap, preserving [Page N] markers. The document_chunks table stores embeddings, and a match_document_chunks RPC does cosine similarity search via pgvector. Indexing fires on document upload; results surface through a /search/semantic endpoint and a semantic_search LLM tool, with a SemanticSearchWidget on ProjectPage.

A few things to check before importing: the match_document_chunks function is security definer and trusts the backend to pass the right p_user_id - no client-side tenant isolation. Indexing is fire-and-forget with no retry or backfill path, so documents uploaded before the migration exists will have no embeddings. Chunking is naive fixed-window; no sentence or paragraph boundary awareness. Migration 001_add_openrouter_voyage.sql requires the vector extension to be enabled in the Supabase dashboard before running.

The two features ship in one commit, which makes cherry-picking awkward. Splitting them on import is worth the effort.

So what Worth a look if you want multi-provider LLM support or semantic search on legal documents. The OpenRouter adapter is a clean drop-in extension. The Voyage embedder choice shows product intent. Skip if fixed-window chunking or fire-and-forget indexing is a problem for your reliability requirements.

Spotted something wrong? Or know the PR text has fresher detail than the writeup above?

Commits in this thread

1 commit from Chris-o-O/mike-intell, oldest first. Source extracted verbatim from the harvested git log.

SHA	Subject	Author	Date
`3279a584`	Add OpenRouter multi-model support and Voyage AI semantic search	Claude	2026-05-07	↗ GitHub
commit body OpenRouter: extends the LLM provider abstraction to support GPT-4o, GPT-4 Turbo, Mistral Large, Llama 3.3 70B, and Gemma 3 27B via a single OpenRouter API key. Follows the same provider pattern as Claude and Gemini - new openrouter.ts adapter, model routing by prefix, per-user API key stored in user_profiles, and full UI integration (model toggle, availability checks, account settings key field). Semantic search: adds Voyage AI embeddings (voyage-law-2, optimised for legal documents) to index document chunks at upload time (fire-and-forget, controlled by VOYAGE_API_KEY env var). Chunks are stored in a new document_chunks pgvector table. A /search/semantic endpoint and a semantic_search LLM tool enable concept-level retrieval. A new SemanticSearchWidget in ProjectPage exposes document search by meaning. https://claude.ai/code/session_01NJxa5NasJ9qhaMkzEyqr2u

SHA

Subject

Author

Date

3279a584

Add OpenRouter multi-model support and Voyage AI semantic search

Claude

2026-05-07

↗ GitHub

commit body

OpenRouter: extends the LLM provider abstraction to support GPT-4o,
GPT-4 Turbo, Mistral Large, Llama 3.3 70B, and Gemma 3 27B via a
single OpenRouter API key. Follows the same provider pattern as Claude
and Gemini - new openrouter.ts adapter, model routing by prefix, per-user
API key stored in user_profiles, and full UI integration (model toggle,
availability checks, account settings key field).

Semantic search: adds Voyage AI embeddings (voyage-law-2, optimised for
legal documents) to index document chunks at upload time (fire-and-forget,
controlled by VOYAGE_API_KEY env var). Chunks are stored in a new
document_chunks pgvector table. A /search/semantic endpoint and a
semantic_search LLM tool enable concept-level retrieval. A new
SemanticSearchWidget in ProjectPage exposes document search by meaning.

https://claude.ai/code/session_01NJxa5NasJ9qhaMkzEyqr2u

Capture this thread into my fork

Download a single Markdown prompt that tells Claude how to port every commit above into your working tree — adapting paths and structure to match your repo. Run it via claude -p < capture-thread-172.md from inside the repo you want the changes in.