Next.js pages router + TanStack Start (mid-migration, see below), React 19. Dev server: pnpm dev:studio → http://localhost:8082. Load the skills matching the task; stack them when a task spans areas: Studio is migrating from the Next.js pages router (pages/) to TanStack Start (routes/). Both runtimes ship side-by-side; the STUDIOFRAMEWORK env var selects which one pnpm dev/build runs (default: next, resolved in scripts/dispatch.js). Full route map and strategy: TANSTACKMIGRATION.md.
Read CLAUDE.md before repository work. Binding routing: this repo owns skills, Engine, and MV3 extension; Agent SDK work belongs in /Users/julb/Desktop/GitHub/caveman-agent-sdk; Browse driver/MCP/benchmark/plugin work belongs in /Users/julb/Desktop/GitHub/caveman-browse; proprietary Pebble product…
Detect a payload's type → route to a safety-classed compressor → count the token reduction → store the original for recovery. The stable four-call API (Compress/Retrieve/Detect/Stats) is what the proxy,…
The consumer top-of-funnel: a toggle that, on ChatGPT / Claude / Gemini, prepends the caveman skill directive to each message you send and re-fires the send, so the AI replies…
A thin stdio JSON-RPC adapter exposing the compression engine as five MCP tools to any host (Claude Code, Cursor, …). It owns only the MCP framing; all compression is the…
Durable, cross-session memory: remember / recall / supersede / history / forget. A local SQLite store holds the raw memories (the source of truth); recall ranks them with deterministic BM25…
A base-URL-swap reverse proxy: match → authenticate → inspect → byte-safe transform → upstream → meter. Single-operator, BYOK, zero cloud dependencies. It shares its provider adapters with the managed gateway…
Compress MCP/OpenAI tool definitions before they fill the context window. A thin Go wrapper over the engine's toolschema compressor: it drops annotation metadata (examples, titles, $comment, $schema) and reduces long…
These rules apply to all public contributions and interactions, no matter how they're done. The entity operating or authorizing the contributions and interactions is responsible for them. AI assistance does not relieve contributors of responsibility for their work. "The AI did it" is not an excuse. Violations may result in contributions being closed without further comment. Cases of deception, evasion, or repeat violations may result in account bans. Allowed: - All AI usage that remains non-public (e.g. questions, reviews, local…