[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-82276":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":13,"subscribersCount":13,"size":13,"stars1d":13,"stars7d":16,"stars30d":17,"stars90d":13,"forks30d":13,"starsTrendScore":13,"compositeScore":13,"rankGlobal":10,"rankLanguage":10,"license":18,"archived":19,"fork":19,"defaultBranch":20,"hasWiki":19,"hasPages":19,"topics":21,"createdAt":10,"pushedAt":10,"updatedAt":30,"readmeContent":31,"aiSummary":32,"trendingCount":13,"starSnapshotCount":13,"syncStatus":33,"lastSyncTime":34,"discoverSource":35},82276,"beevibe-cto","beevibe-ai\u002Fbeevibe-cto","beevibe-ai","Architecture Deep Research: deep research for strategic system design decisions.","https:\u002F\u002Fbeevibe.ai\u002Fcto",null,"JavaScript",104,0,37,1,17,60,"Apache License 2.0",false,"main",[22,23,24,25,26,27,28,29],"agent-skills","ai","ai-agents","cli","cto","deep-research","mcp","system-design","2026-06-12 02:04:24","# Beevibe AI CTO\n\n**The decision layer your coding agents are missing.**\n\n\u003Cimg alt=\"Beevibe AI CTO\" src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F4f8dd109-71de-4cfa-b957-b609bf50591a\" width=\"100%\" \u002F>\n\n\u003Cdetails>\n\u003Csummary>\u003Cb>More screenshots\u003C\u002Fb>\u003C\u002Fsummary>\n\u003Cbr>\n\n\u003Cimg alt=\"Beevibe AI CTO\" src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002Fd3f2f4ce-7c00-4d21-8a7e-824e871e5138\" width=\"100%\" \u002F>\n\u003Cimg alt=\"Beevibe AI CTO\" src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002Ff07a64e7-563b-47a2-bba1-399215358f85\" width=\"100%\" \u002F>\n\u003Cimg alt=\"Beevibe AI CTO\" src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F12b8a8a5-850c-4e4d-81c1-12b5bd7525d9\" width=\"100%\" \u002F>\n\u003Cimg alt=\"Beevibe AI CTO\" src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002Fcec470bb-0411-45a1-b6e7-6029d82bcb33\" width=\"100%\" \u002F>\n\u003Cimg alt=\"Beevibe AI CTO\" src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002Ff9ca2f06-4ebe-48f9-96f0-852b0632b149\" width=\"100%\" \u002F>\n\u003Cimg alt=\"Beevibe AI CTO\" src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002Fa09f45bd-9761-441c-83dd-1eb11c29f414\" width=\"100%\" \u002F>\n\n\u003C\u002Fdetails>\n\nThe full loop, from architecture decision through PR review to drift detection:\n\n| Command | What it does |\n| --- | --- |\n| `adr decide` | Live deep-research on an architectural decision → an HTML report with N candidates, citations, and Mermaid diagrams. **[See an example →](https:\u002F\u002Fbeevibe.ai\u002Fcto\u002Fexample-report\u002F)** |\n| `adr principles init` | Scans your repo, discovers the team's code-review lenses (state-boundaries, schema-validate-before-write, etc.), walks you through an interview, writes `.adr\u002Fprinciples.{md,json}` |\n| `adr principles refresh` | Re-discover without losing the interview log. Asks only about new ambiguities. |\n| `adr principles incremental` | Only re-extract from files changed since last refresh. Reuses prior lenses. Fast. |\n| `adr principles refine --id \u003Cid>` | Single-principle re-discovery for one specific rule. |\n| `adr review` | Checks a PR \u002F staged diff \u002F branch against the principles. Walks the user through violations one-by-one, posts inline comments via `gh` \u002F `glab` \u002F Bitbucket API. Supports `\u002F\u002F adr-ignore: \u003Cid>` suppression and `--batch` for one-call CI reviews. |\n| `adr drift` | Full-repo scan vs principles. Useful for \"how far has the codebase drifted since principles init six months ago?\". |\n| `adr guard install` | Wires the principles into Claude Code (PreToolUse hook surfaces them at write time) + git pre-commit (blocks high-severity violations). `adr guard uninstall` reverses it. |\n\nThe principles file evolves automatically: stale citations get flagged, accept\u002Fedit\u002Fskip stats get persisted, and confidence demotes principles users keep skipping. **No static-lint trap.** See [ROADMAP.md](.\u002FROADMAP.md) for the evolvability story and the commit history.\n\nThe brain closes the rest of the loop; upcoming.\n\n---\n\n## Install\n\nThree ways in. All share the same kernel.\n\n### Claude Code plugin — recommended\n\n```bash\nclaude plugin marketplace add beevibe-ai\u002Fbeevibe-cto\nclaude plugin install adr\n```\n\nThen in any Claude Code session:\n\n| Slash | What it does |\n| --- | --- |\n| `\u002Fadr:doctor` | One-time: audit env, walk through API keys, persist to `~\u002F.adr\u002Fconfig.json` |\n| `\u002Fadr:decide` | Ask a decision name, scan the repo, run the full pipeline, summarize the report |\n| `\u002Fadr:discover` | Quick scan only — drafts a PRD without the full deep-research run |\n| `\u002Fadr:principles` | Discover the team's code-review principles from the repo, walk through the interview conversationally |\n| `\u002Fadr:review` | Check a PR \u002F staged diff against the principles, walk through violations one at a time |\n| `\u002Fadr:guard` | Install the PreToolUse + pre-commit hooks so principles fire automatically on every edit |\n\n### MCP server — Cursor, Codex, any MCP host\n\n```bash\nnpm install -g github:beevibe-ai\u002Fbeevibe-cto\nadr-doctor setup\n```\n\nAdd to your host's MCP config:\n\n```json\n{ \"mcpServers\": { \"adr\": { \"command\": \"adr-mcp\" } } }\n```\n\nFive tools become available: `adr_discover`, `adr_deep_research`, `adr_read_handoff`, `adr_principles`, `adr_review`.\n\n### CLI — terminal, CI, GitHub Action\n\n```bash\nnpm install -g github:beevibe-ai\u002Fbeevibe-cto\nadr-doctor                       # audit env, exit non-zero if anything missing\n```\n\n## Run your first decision\n\n```bash\nadr deep-research --discover-first --include-peers --open \\\n  --repo . \\\n  --domain \"your product domain\" \\\n  --decision \"vector store for agent memory\" \\\n  --out .adr-runs\u002Fvector-store\n```\n\nWhat this does:\n\n1. **Discover** scans your repo for stack signals, patterns the team follows, and antipatterns the team has explicitly rejected.\n2. **Peer-finder** names 3-5 similar products. Open-source peers (Neo4j, Onyx) get read through their repos and engineering blogs. Closed-source peers (Notion, Obsidian, Mem.ai) get read through Reddit, HN, Twitter — what users actually report.\n3. **Research** collects live evidence, builds a comparison matrix, runs adversarial probes against every candidate.\n4. **Synthesis** writes the research report. Citation + claim audits run automatically.\n5. **`--open`** renders `ADR.md` as HTML (mermaid diagrams as SVG, tables, dark\u002Flight mode) and opens it in your default browser.\n\nA typical run takes 3-6 minutes and costs $0.10-$0.30 in API spend. Use `--dry-run` to see the plan + cost estimate without spending tokens.\n\n## After the report\n\n```bash\n# Open the report later — or after a run that didn't use --open\nadr open .adr-runs\u002Fvector-store\n\n# Pick an option from the report and generate its implementation contract\nadr handoff .adr-runs\u002Fvector-store --option pgvector\n\n# Resume a crashed or interrupted run (reuses cached evidence.json — the expensive part)\nadr resume .adr-runs\u002Fvector-store\n```\n\nThe report at `\u003Cout_dir>\u002FADR.md` has:\n\n- Executive Summary + Option Space table\n- One section per candidate: evidence depth (`thick` \u002F `medium` \u002F `thin`), what the evidence shows, what it doesn't, pick-when \u002F avoid-when reading aids, citations\n- Cross-Cutting Tradeoffs across matrix axes\n- Open Questions the evidence pool didn't resolve\n- Where to Dig Deeper — pre-filled `adr deep-research` commands for the next iteration\n\nThe decision becomes a tree of ADR runs. Each one drills into the highest-uncertainty axis from the prior run.\n\n## Keep the implementation honest — principles, review, guard\n\nDecisions don't survive contact with the codebase. The senior engineer who knows what your team standardized on becomes the only memory — repeating the same review comment to 10 different juniors, every week.\n\nThree commands close that loop:\n\n```bash\n# Step 1: discover what the team's conventions actually are (run once per repo)\nadr principles init                    # scans the repo, interviews you, writes .adr\u002Fprinciples.{md,json}\n\n# Step 2: check a PR against those principles\nadr review 42                          # GitHub PR\nadr review --staged                    # pre-commit local\nadr review --branch main               # current branch vs base\n\n# Step 3: wire the principles into your workflow so they fire automatically\nadr guard install                      # PreToolUse hook + git pre-commit\n```\n\n**`adr principles init`** uses the LLM end-to-end — no hardcoded lens taxonomy. It samples real source files from your repo, asks the model to surface the code-review angles a senior would catch (state boundaries, schema-validate-before-write, CLI patterns, test-fixture discipline, ownership routing…), extracts positive patterns + antipatterns + ambiguities with `file:line` citations, then walks you through an interactive interview to confirm. A cite-or-die filter drops any path the model invented — only principles backed by real lines on disk survive.\n\n**`adr review`** loads the principles + a unified diff (PR \u002F staged \u002F branch \u002F arbitrary diff file \u002F stdin), groups hunks by file, runs one LLM call per file to detect violations, ranks by severity, and walks you through accept\u002Fedit\u002Fskip one at a time. Approved comments post via `gh pr review` with the team's own example cited as \"follow this\".\n\n**`adr guard install`** wires two hooks:\n- **Claude Code `PreToolUse`** — every Edit\u002FWrite\u002FMultiEdit, the hook filters principles to ones relevant to the file's top-level dir (no LLM call on the hot path) and injects them into the agent's context. The agent sees the team's conventions before it generates the violation.\n- **Git pre-commit** — runs `adr review --staged --top-n 5`. HIGH-severity violations block the commit; medium\u002Flow are advisory.\n\nAll three operate on the same `.adr\u002Fprinciples.json` artifact. Re-run `adr principles init` whenever the team's posture shifts; review and guard pick up the new rules automatically.\n\n## API keys\n\n`adr-doctor setup` walks you through these interactively and stores them in `~\u002F.adr\u002Fconfig.json` (mode 0600). Process env always overrides the file.\n\nRequired (at least one of each):\n\n| Group | Env var | Free tier |\n| --- | --- | --- |\n| Search | `BRAVE_SEARCH_API_KEY` | ~2k queries\u002Fmo, https:\u002F\u002Fapi-dashboard.search.brave.com |\n| Search | `TAVILY_API_KEY` | 1k requests\u002Fmo, https:\u002F\u002Ftavily.com |\n| Search | `SERPER_API_KEY` | 2.5k queries on signup, https:\u002F\u002Fserper.dev |\n| Search | `SEARXNG_URL` | self-hosted, https:\u002F\u002Fdocs.searxng.org |\n| LLM | `ADR_OPENAI_API_KEY` (or `OPENAI_API_KEY`) | https:\u002F\u002Fplatform.openai.com\u002Fapi-keys |\n\nIf no dedicated search key is set, ADR falls back to OpenAI's hosted `web_search` — one key powers both research and synthesis.\n\nOptional:\n\n- `GITHUB_TOKEN` — strongly recommended. Lifts the GitHub API limit from 60\u002Fhr to 5000\u002Fhr.\n- `ADR_MODEL` — override the default model (`gpt-4.1-mini`).\n- `ADR_OPENAI_BASE_URL` — point at a local OpenAI-compatible server (vLLM, LM Studio, llamafile, Ollama).\n- `ADR_SEARCH_INCLUDE_DOMAINS` \u002F `ADR_SEARCH_EXCLUDE_DOMAINS` — bias the evidence pool toward \u002F away from specific domains.\n- `ADR_MCP_SERVER_URL` + `ADR_PRIVATE_MCP_ONLY` — search a read-only private MCP corpus instead of the public web.\n\n## What ADR produces\n\nA run's `\u003Cout_dir>\u002F` contains:\n\n| File | What it is |\n| --- | --- |\n| `ADR.md` \u002F `ADR.html` | Reader-facing research report (HTML is generated by `adr open`). |\n| `research-report.json` | Structured report — same content, machine-readable. |\n| `comparison-matrix.json` | Candidates × axes table that feeds the report. |\n| `evidence.json` + `source-snapshots\u002F` | Audit trail. Every claim cites a snapshot. |\n| `peers.json` | Peers surfaced by `--include-peers`, with `evidence_strategy` per peer. Editable. |\n| `decision-context.json` | Context annotations extracted from your PRD. Editable. |\n| `follow-up-questions.json` | Pre-filled `adr deep-research` commands for the highest-spread axes. |\n| `state.json` \u002F `cost.json` \u002F `events.jsonl` | Run lifecycle, cost ledger, live event log. |\n\nAfter `adr handoff --option \u003Cname>`:\n\n| File | What it is |\n| --- | --- |\n| `agent-guardrails.md` | Implementation contract for the chosen candidate. |\n| `execution-handoff.json` | Structured handoff for downstream coding agents. |\n\n## Verify your install\n\n```bash\nnpm test\n```\n\nSix suites run locally — kernel regression, search provider, schema check, framework + web + MCP smoke. No network calls; green here means the wiring is intact.\n\nTo exercise the live loop:\n\n```bash\nadr-doctor                       # confirm READY\nadr deep-research --discover-first --open \\\n  --repo . --domain \"test\" --decision \"retrieval topology\" \\\n  --out .adr-runs\u002Fself-test\n```\n\n## Status\n\n**Shipped:**\n\n- **ADR flagship** — `adr decide`, `adr discover`, `adr open`, `adr handoff`, `adr resume`, `adr supersede`. Live agentic research kernel, peer discovery, community sources, citation + claim audits, Mermaid diagrams, HTML report, crash-aware state.\n- **`adr principles`** — `init` \u002F `refresh` \u002F `incremental` \u002F `refine`. LLM-discovered code-review lenses, per-lens patterns + antipatterns, interactive interview, cite-or-die filter. Cite-rot detection on every review; accept\u002Fedit\u002Fskip stats persist; confidence auto-evolves from how the team actually uses each rule.\n- **`adr review`** — PR \u002F staged \u002F branch \u002F file modes. Walks the user through violations one at a time, posts inline comments via `gh` (GitHub), `glab` (GitLab), or Bitbucket REST API. Supports `\u002F\u002F adr-ignore: \u003Cprinciple-id>` suppression. `--batch` flag for one-call CI reviews.\n- **`adr drift`** — full-repo scan vs principles, parallel-bounded. Useful for periodic \"how far has the code drifted?\" audits.\n- **`adr guard`** — `install` \u002F `uninstall`. Claude Code `PreToolUse` hook (write-time principle injection) + git pre-commit (blocks high-severity violations). Pure file-based filter on the hot path; idempotent.\n- **Adapters + UI** — LangGraph and Google ADK (same kernel, same artifacts). Web UI (`adr-web`) for live operator \u002F developer views.\n- **Claude Code plugin** — six slash commands: `\u002Fadr:doctor`, `\u002Fadr:decide`, `\u002Fadr:discover`, `\u002Fadr:principles`, `\u002Fadr:review`, `\u002Fadr:guard`.\n\n**In development:**\n\n- **The brain** — always-on knowledge graph that watches voices, trending OSS, competitor architecture, and papers. Personalized to your stack via your PRD + past ADR runs + discovered principles. Markdown-in-git source of truth (Obsidian-compatible) + derived Postgres+pgvector indexes; LLM maintains the wiki, watchers ingest sources continuously.\n\nOpen-source core under Apache-2.0. The commercial Beevibe surface layers curated corpora, managed researcher agents, org-level memory, and team governance on top.\n\n## Learn more\n\n- **[See an example report](https:\u002F\u002Fbeevibe.ai\u002Fcto\u002Fexample-report\u002F)** — real run on a Beevibe decision, rendered the same way `adr open` would render yours.\n- [ADR introduction](https:\u002F\u002Fbeevibe.ai\u002Fblog\u002F03-beevibe-cto\u002F) — the layer before the coding agent.\n- [Questions teams keep asking](https:\u002F\u002Fbeevibe.ai\u002Fblog\u002F04-adr-questions\u002F) — Q&A on the design.\n- [The dogfooding journey](https:\u002F\u002Fbeevibe.ai\u002Fblog\u002F06-after-dogfooding\u002F) — what made us pivot from a decision engine to a research-report engine.\n- [docs\u002F](.\u002Fdocs\u002F) — framework adapters, web UI, schemas, mesh integration, full flag reference.\n","Beevibe AI CTO 是一个专注于系统架构决策的深度研究工具。它通过一系列命令帮助开发者从架构决策到代码审查再到漂移检测，实现全生命周期管理。核心功能包括生成详细的架构决策报告、初始化和刷新团队代码审查原则、检查代码变更是否符合原则以及扫描整个仓库以检测代码漂移。该工具使用JavaScript编写，支持多种CI\u002FCD集成方式，并且具有自动更新原则文件的能力，确保其始终与项目需求保持一致。适用于需要进行复杂系统设计和技术选型的软件开发团队，特别是那些希望提高代码质量和一致性、减少技术债务的团队。",2,"2026-06-11 04:08:14","CREATED_QUERY"]