improve: add README, .editorconfig, fix fallback models in setup_all.py, update .gitignore

.editorconfig

@@ -0,0 +1,26 @@
+# EditorConfig helps maintain consistent coding styles across editors and IDEs
+# https://editorconfig.org
+
+root = true
+
+[*]
+charset = utf-8
+end_of_line = lf
+insert_final_newline = true
+trim_trailing_whitespace = true
+max_line_length = 100
+
+[*.{py,pyi}]
+indent_style = space
+indent_size = 4
+
+[Makefile]
+indent_style = tab
+
+[*.{nix}]
+indent_style = space
+indent_size = 2
+
+[*.md]
+trim_trailing_whitespace = false
+max_line_length = 0

.gitignore

@@ -10,3 +10,8 @@ whsec*.txt
 __pycache__/
 *.pyc
 result
+*.egg-info/
+.eggs/
+dist/
+build/
+.venv/

README.md

@@ -0,0 +1,55 @@
+# PR-Agent Server
+
+Nix-deployed GitHub App server for automated PR review + auto-merge, using a custom LLM endpoint (9router/Omniroute).
+
+## Architecture
+
+```
+GitHub webhook → Cloudflare DNS → Caddy (reverse proxy, :4002)
+    → pr-agent-server (Nix profile, uvicorn)
+        → PR-Agent github_app.py (FastAPI)
+            → 9router API (custom OpenAI-compatible endpoint)
+```
+
+## Components
+
+| File | Purpose |
+|------|---------|
+| `run_server.py` | Main FastAPI server — GitHub App webhook handler + analytics/metrics + Discord notifications |
+| `auto_merge_bot.py` | Periodic bot that finds reviewed PRs and merges them |
+| `trivial_merge.py` | Trivial PR fast-path (docs-only, dependabot, <100 lines) |
+| `health-check.py` | Model health watchdog — tests primary + fallback models against 9router |
+| `sync-key.py` | Auto-syncs the BWS router key to disk on every service start |
+| `generate_manifest.py` | Creates GitHub App manifest URL |
+| `callback_server.py` | Dev callback server for receiving GitHub App credentials |
+| `flake.nix` | Nix build definition — produces the deployable package |
+
+## Development
+
+```bash
+# Syntax check all Python files
+python3 -m py_compile run_server.py auto_merge_bot.py health-check.py sync-key.py trivial_merge.py callback_server.py generate_manifest.py setup_all.py setup_app.py start_server.py
+
+# Build with Nix
+nix build .#default
+
+# Deploy (CI does this automatically on push to main)
+nix copy --to ssh://user@vps $STORE_PATH
+ssh user@vps "sudo /nix/var/nix/profiles/default/bin/nix-env --profile /nix/var/nix/profiles/pr-agent-server --set '$STORE_PATH'"
+ssh user@vps "sudo systemctl restart pr-agent-server"
+```
+
+## Ops
+
+- **Health watchdog**: cron `pr-agent-health-watchdog` (every 10m) → `~/.hermes/scripts/pr-agent-health-check.sh`
+- **Key sync**: systemd `ExecStartPre` → `pr-agent-sync-key` (BWS → disk)
+- **Prometheus**: `GET /api/metrics` → `pr_agent_requests_total`, `pr_agent_requests_by_command`, `pr_agent_model_failures`
+- **Analytics**: `GET /api/analytics` → JSON summary of recent events + failures
+- **Discord**: `POST /api/v1/notify_review` → webhook delivery for review complete/failed
+
+## Nix Profile Integrity
+
+After deploy, run `nix-gc-vps.sh` to clean up old generations. The GC script now includes
+a repair step that fixes broken profile symlinks before running `nix store gc`, preventing
+the issue where profile `-link` dirs get deleted and store paths become unreferenced (see
+`devops/pr-agent-deployment` skill for full troubleshooting).

setup_all.py

@@ -78,8 +78,8 @@ with open(BASE_DIR / ".secrets.toml", "w") as f:
 
 # ── 4. Create PR-Agent config ──
 config_toml = """[config]
-model = "claude-opus-4-8"
+model = "openai/claude-opus-4-8"
-fallback_models = ["auto/best-coding", "auto/claude-sonnet"]
+fallback_models = ["openai/ATLAS", "openai/gemini", "openai/text", "openai/deepseek-v4-free"]
 custom_model_max_tokens = 128000
 git_provider = "github"
 publish_output = true