hermes-context-searchlisted

# hermes-context: multi-repo semantic search (fallback — prefer `grepai` MCP tool) **Use `grepai` instead when available** (check `.mcp.json` for the `grepai` MCP server, or run `grepai status` — see `docs/LOCAL-SEMANTIC-CODE-INDEX.md`). Both use the same local `nomic-embed-text` Ollama model at $0 cost; grepai's AST-aware chunking and native MCP integration (a first-class tool call, not a shell-out) make it the better default. Fall back to this tool only if `grepai` isn't installed/configured in the current repo. ## When to invoke - A task touches a repo you have not explored yet this session, and a targeted grep query isn't obvious (concept search, not string search). - A question spans MULTIPLE repos ("where else in the fleet do we handle gateway fallback routing") — `hermes-context search` queries every indexed repo at once, including ones not currently checked out in this session. - Do NOT invoke for a single known file/symbol lookup, or a repo already loaded into context this session — Grep/Glob are faster and free of the (still real, un-optimized) embedding-index latency below. ## Commands ``` hermes-context index /path/to/repo [--name X] # (re)build the index; incremental — unchanged files are never re-embedded hermes-context search "concept or question" [--repo X] [--top N] [--json] hermes-context list # which repos are indexed, chunk counts hermes-context doctor [--json] # ollama up? model pulled?