Feat/GitHub copilot provider

PR_REVIEW_PLAN.md

@@ -0,0 +1,81 @@
+# PR Review Completion Plan
+
+Goal: make `okcode` PR Review feel comprehensive, reliable, and maintainer-grade on the existing `/_chat/pr-review` surface.
+
+## Phase 1 — Finish the current review cockpit
+
+### 1. Action rail polish
+
+- [ ] Tighten the action rail layout so decision, blockers, file impact, and recent maintainer reviews read cleanly at a glance.
+- [ ] Make approval blockers explicit: conflicts, failing checks, pending checks, blocked workflow steps.
+- [ ] Ensure copy is compact and maintainer-oriented.
+
+### 2. Review state clarity
+
+- [ ] Show the current PR review decision clearly and consistently.
+- [ ] Confirm the latest submitted review state refreshes correctly after comment / approve / request changes.
+- [ ] Verify draft review body reset and query invalidation behavior after submit.
+
+### 3. Review history quality
+
+- [ ] Improve recent maintainer reviews presentation for scanability.
+- [ ] Distinguish maintainer review state from plain discussion/comment noise.
+- [ ] Decide whether the compact summary is enough or needs a deeper expandable history view.
+
+## Phase 2 — Deepen review context
+
+### 4. File-level review signal
+
+- [ ] Surface clearer per-file review context: commented files, unresolved-thread files, reviewed files.
+- [ ] Make it easier to see where maintainer attention is still needed.
+- [ ] Verify selected-file behavior stays stable when data refreshes.
+
+### 5. Blockers and workflow visibility
+
+- [ ] Make mergeability, required checks, conflicts, and workflow blockers easier to understand from the page.
+- [ ] Improve “why can’t I approve yet?” guidance.
+- [ ] Ensure blocker state is visible without needing to hunt through inspector panels.
+
+### 6. Maintainer workflow integration
+
+- [ ] Confirm PR Review is fully discoverable through navigation and command palette.
+- [ ] Identify any remaining maintainer entry points that should route into `/_chat/pr-review`.
+- [ ] Avoid creating duplicate surfaces or split workflows.
+
+## Phase 3 — End-to-end reliability
+
+### 7. Validation and cleanup
+
+- [ ] Run full validation for the changed slice, including typecheck if possible.
+- [ ] Fix any lint/type/test issues introduced by the PR Review work.
+- [ ] Keep scope tight and avoid unrelated file churn.
+
+### 8. Real PR walkthrough
+
+- [ ] Test the full maintainer flow against a real PR.
+- [ ] Verify open review, inspect files, inspect threads, view recent maintainer reviews, submit review, and refresh behavior.
+- [ ] Capture any UX or state-sync gaps found in real usage.
+
+### 9. Final completeness pass
+
+- [ ] Review the page as a whole for maintainer usability.
+- [ ] Confirm the implementation feels like the canonical PR Review workflow, not a partial bolt-on.
+- [ ] Produce a final summary of what is complete vs what is intentionally deferred.
+
+## Execution order
+
+1. Action rail polish
+2. Review state clarity
+3. Review history quality
+4. File-level review signal
+5. Blockers and workflow visibility
+6. Maintainer workflow integration
+7. Validation and cleanup
+8. Real PR walkthrough
+9. Final completeness pass
+
+## Definition of done
+
+- A maintainer can open `/_chat/pr-review`, understand PR state quickly, inspect review context, see recent maintainer activity, submit a review confidently, and trust the page to stay in sync afterward.
+- The page is clearly the main PR review workflow in `okcode`.
+- Validation is clean enough that we trust the implementation, not just the visuals.

apps/server/package.json

@@ -26,6 +26,7 @@
     "@anthropic-ai/claude-agent-sdk": "^0.2.77",
     "@effect/platform-node": "catalog:",
     "@effect/sql-sqlite-bun": "catalog:",
+    "@github/copilot-sdk": "^0.2.1",
     "@pierre/diffs": "^1.1.0-beta.16",
     "effect": "catalog:",
     "node-pty": "^1.1.0",

apps/server/src/doctor.ts

@@ -10,6 +10,7 @@ import { Effect } from "effect";
 import { Command } from "effect/unstable/cli";
 
 import {
+  checkCopilotProviderStatus,
   checkCodexProviderStatus,
   checkClaudeProviderStatus,
 } from "./provider/Layers/ProviderHealth";
@@ -32,6 +33,7 @@ const AUTH_LABELS: Record<string, string> = {
 const PROVIDER_LABELS: Record<string, string> = {
   codex: "Codex (OpenAI)",
   claudeAgent: "Claude Code",
+  copilot: "GitHub Copilot",
 };
 
 function printStatus(status: ServerProviderStatus): void {
@@ -65,9 +67,12 @@ const doctorProgram = Effect.gen(function* () {
   console.log("");
   console.log("Checking provider health...");
 
-  const statuses = yield* Effect.all([checkCodexProviderStatus, checkClaudeProviderStatus], {
-    concurrency: "unbounded",
-  });
+  const statuses = yield* Effect.all(
+    [checkCodexProviderStatus, checkClaudeProviderStatus, checkCopilotProviderStatus],
+    {
+      concurrency: "unbounded",
+    },
+  );
 
   for (const status of statuses) {
     printStatus(status);
@@ -80,6 +85,7 @@ const doctorProgram = Effect.gen(function* () {
     console.log("");
     console.log("  Codex:  npm install -g @openai/codex && codex login");
     console.log("  Claude Code: npm install -g @anthropic-ai/claude-code && claude auth login");
+    console.log("  Copilot: npm install -g @github/copilot && copilot login");
   } else if (readyCount === statuses.length) {
     console.log("All providers are ready.");
   } else {