云服务部-叶林立andDevin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
f9a7f743dd
- codemode：Tool Search 常驻（alwaysEnabled）、权限 allow（嵌套调用仍逐个过权限门）、 tools.jsonc 登记使用规则；RTK 跳过 parentToolCallId 嵌套调用，避免丢失 bash structuredContent；Tool Search 目录过滤 hidden / codemode（script-only）exposure； 扩展加载测试加载 builtin:codemode 并断言其常驻激活 - 搜索 key 不再经命令行参数传递：search-auth.mjs 改为 --set-stdin（先读校验再加锁）， configure.sh 交互输入经管道、非交互用 --<provider> - - CodeGraph 客户端监听子进程 stdin error，避免 EPIPE 未处理导致崩溃 - 移除 PI_TOOL_SEARCH/HERMES_MEMORY/RTK_CONFIG_PATH 环境变量入口，测试改用工厂 configPath 注入 - pi-subagents 固定 vitest / coverage-istanbul 4.1.10，修复 npm ls 版本冲突 - 补 pi-web-search / pi-worktree UPSTREAM.md 与 pi-minimal-footer AGENTS.md， files 白名单补齐并加测试；.codegraph/.gitignore 移出跟踪 - detect_kitty 优先识别官方安装目录 Generated with [Devin](https://devin.ai) Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
pi-web-search
Unified web search and fetch for Pi. The main model only ever sees two tools —
web_search and web_fetch — provider selection happens inside the extension.
web_search(query, max_results?, site?, published_after?, published_before?, depth?)routes each query to Tavily, Exa, or Keenable using the session's current main model (ctx.model+ctx.modelRegistry.complete), guided by per-provider rules.max_resultsdefaults to 8 and is clamped to 1–20. Results includedetails.provider/details.routedByfor debugging.web_fetch(urls, max_characters?)always reads pages via Exa (web_fetch_exa), degrading silently to Keenable then Tavily.max_charactersdefaults to 20,000 and is clamped to 1–100,000 per page; oversized fetch output is further bounded by pi-overflow downstream.
Providers (clients/)
Plain fetch calls — no MCP adapter, no upstream extension wrappers:
clients/tavily.ts—POST api.tavily.com/search|extract, Bearer key. Requires atavilycredential.clients/keenable.ts—api.keenable.ai, keyless/publicendpoints by default, akeenablecredential raises rate limits. Adapted from@keenable/pi-search.clients/exa.ts— JSON-RPC overmcp.exa.ai/mcp(Streamable HTTP):initializeonce, cachemcp-session-id,tools/callper request, re-init on session expiry. Anexacredential is optional (x-api-keyheader).
Keys live in Pi's auth.json as one api_key credential per provider
(tavily, exa, keenable), read by auth.ts with upstream resolution
semantics — literal values, $VAR/${VAR} interpolation, and !command
indirection (e.g. "!op read op://Vault/Tavily/key"). configure.sh search
writes them via its hidden interactive prompt or scripts/search-auth.mjs --set-stdin reading the key from stdin — a key is never accepted as a
command-line argument (atomically, mode 600, preserving all other entries);
process
environment variables are not read directly — use "key": "$TAVILY_API_KEY"
to indirect through env. KEENABLE_API_URL stays env-only, or can be stored
in the keenable credential's env map.
Routing
router.ts asks the current model to answer with exactly one provider name,
given a prompt containing the per-provider strengths and the query's requested
filters. Decisions are cached per session (LRU, 32 entries). Failure paths are
deterministic: an unroutable answer or a dead provider falls back across the
remaining available providers — never a heuristic choice over the model.