No longer active. This fork has been removed from GitHub. The content below is preserved from MikeWatch's archive.

Chris-o-O/mike-intell

Multi-provider LLM access, legal-domain semantic search, and firm-deployable persona injection

Two active development threads. Provider-neutrality: OpenRouter adapter plus Voyage AI semantic search, shipped together in one commit. Multi-tenancy: org tables, RLS policies, a 444-line organizations router, and system-prompt persona injection across chat and project chat.

View on GitHub →

Chris-o-O/mike-intell adds two clusters of changes on top of the upstream codebase. The first extends the LLM layer with an OpenRouter adapter (GPT-4o, Mistral Large, Llama 3.3 70B, Gemma 3 27B) and bolts on pgvector-backed semantic search using Voyage AI's voyage-law-2 embedding model. The second introduces a full multi-tenant persona layer: organizations, member roles, email-based invites, and a persona.ts module that injects org and user context into every system prompt.

The fork's trajectory points toward firm-level deployment. First-class support for jurisdictions, practice areas, and bar numbers - combined with en/fr language flags - suggests a Canadian or bilingual legal market target.

What's in it

Direction

multi-tenantsearchpersonas

Activity

Themed changes and pull requests touching this fork, newest first. Themed changes that haven't been turned into a public post yet still appear — they're real work even without a published writeup.

📝 Chris-o-O gives Mike a firm to answer to 2 commits 2mo ago personasmulti-tenant draft
The assistant stops being a generic chatbot and starts knowing which firm, jurisdiction, and practice area it's working for.

Threads of work (detailed view)

4 threads have been distilled into posts.

Chris-o-O gives Mike a firm to answer to

The assistant stops being a generic chatbot and starts knowing which firm, jurisdiction, and practice area it's working for.

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`.