feat: add AI-powered GitHub repository features

.gitignore

@@ -4,4 +4,5 @@ backend/.env
 backend/data/
 __pycache__/
 *.pyc
-frontend/dist/
+frontend/dist/
+agents/

backend/app/ai/context.py

@@ -5,6 +5,8 @@
 
 SYSTEM_PROMPT = """You are GitHub Atlas, a precise repository intelligence assistant.
 Answer only from the supplied GitHub context. Never invent files, metrics, technologies, or code.
+Repository descriptions are weak evidence and are never enough by themselves to claim a stack or framework. To support a claim, use explicit evidence from the repository metadata, dependency lists, technology detection, README content, source file analysis, and retrieved code snippets.
+If the repository context does not show direct evidence, say that the repository does not provide enough evidence to confirm it.
 When citing evidence, use repository-relative file paths in backticks. Explain uncertainty when context is incomplete.
 For statistics and profile questions, calculate from the supplied structured data.
 For code questions, distinguish static-analysis findings from direct source excerpts.
@@ -51,7 +53,7 @@ def repository_context(analysis: dict, chunks: list[dict]):
                 "chunk_index": chunk["chunk_index"],
                 "content": chunk["content"][:2400]
             }
-            for chunk in chunks[:2]
+            for chunk in chunks[:4]
         ]
     }
 

backend/app/analyzers/commit_analyzer.py

@@ -2,7 +2,7 @@
 from datetime import datetime
 
 
 def analyze_commits(commits: list):
 
     if not commits:
         return {
@@ -10,34 +10,39 @@
             "contributors": [],
             "commits_by_contributor": {},
             "first_commit": None,
-            "latest_commit": None
+            "latest_commit": None,
+            "history": []
         }
 
     contributor_counter = Counter()
-
     commit_dates = []
+    history = []
 
     for commit in commits:
+        author = commit.get("author") or {}
+        username = author.get("login") or commit.get("commit", {}).get("author", {}).get("name")
 
-        author = commit.get("author")
-
-        if author:
-            username = author.get("login")
-
-            if username:
-                contributor_counter[username] += 1
+        if username:
+            contributor_counter[username] += 1
 
         commit_info = commit.get("commit", {})
-        author_info = commit_info.get("author")
-
-        if author_info:
-
-            date = author_info.get("date")
-
-            if date:
-                commit_dates.append(date)
+        author_info = commit_info.get("author") or {}
+        committer_info = commit_info.get("committer") or {}
+        date = author_info.get("date") or committer_info.get("date")
+        if date:
+            commit_dates.append(date)
+
+        history.append({
+            "sha": commit.get("sha"),
+            "message": (commit_info.get("message") or "").strip().split("\n")[0],
+            "author": username or commit_info.get("author", {}).get("name") or "Unknown",
+            "date": date,
+            "url": commit.get("html_url") or commit.get("url"),
+            "committer": (committer_info.get("name") or "Unknown")
+        })
 
     commit_dates.sort()
+    history.sort(key=lambda item: item["date"] or "", reverse=True)
 
     return {
         "total_analyzed": len(commits),
@@ -65,5 +70,7 @@
             commit_dates[-1]
             if commit_dates
             else None
-        )
+        ),
+
+        "history": history
     }

backend/app/routes/auth.py

@@ -40,7 +40,22 @@
 
 
 @router.get("/github/callback")
-async def github_callback(code: str):
+async def github_callback(
+    code: str | None = None,
+    error: str | None = None,
+    error_description: str | None = None,
+):
+    if error:
+        raise HTTPException(
+            status_code=400,
+            detail=(error_description or error or "GitHub authentication failed")
+        )
+
+    if not code:
+        raise HTTPException(
+            status_code=400,
+            detail="Missing GitHub OAuth code"
+        )
 
     token_data = await exchange_code_for_token(code)
 
@@ -63,7 +78,8 @@
         value=session,
         httponly=True,
         secure=False,  # True in production HTTPS
-        samesite="lax"
+        samesite="lax",
+        path="/"
     )
 
     return response

backend/app/services/github_api.py

@@ -1,6 +1,7 @@
 import asyncio
 import httpx
 import base64
+from urllib.parse import quote
 
 GITHUB_API_URL = "https://api.github.com"
 
@@ -170,20 +171,35 @@ async def get_repository_tree(
         "X-GitHub-Api-Version": "2022-11-28"
     }
 
-    async with httpx.AsyncClient() as client:
-
-        response = await client.get(
-            f"{GITHUB_API_URL}/repos/"
-            f"{owner}/{repo}/git/trees/{branch}",
-            headers=headers,
-            params={
-                "recursive": "1"
-            }
-        )
-
-        response.raise_for_status()
+    branches_to_try = [branch, "main", "master"]
+    seen = set()
 
-        return response.json()
+    async with httpx.AsyncClient() as client:
+        for candidate in branches_to_try:
+            if not candidate or candidate in seen:
+                continue
+            seen.add(candidate)
+
+            encoded_candidate = quote(candidate, safe="")
+
+            try:
+                response = await client.get(
+                    f"{GITHUB_API_URL}/repos/"
+                    f"{owner}/{repo}/git/trees/{encoded_candidate}",
+                    headers=headers,
+                    params={
+                        "recursive": "1"
+                    }
+                )
+
+                if response.status_code == 404:
+                    continue
+                response.raise_for_status()
+                return response.json()
+            except httpx.HTTPStatusError:
+                continue
+
+    return {"tree": []}
 
 async def get_repository_file(
     access_token: str,

backend/tests/test_auth_callback.py

@@ -0,0 +1,60 @@
+from fastapi.testclient import TestClient
+
+from app.ai.context import SYSTEM_PROMPT
+from app.analyzers.commit_analyzer import analyze_commits
+from app.main import app
+
+
+client = TestClient(app)
+
+
+def test_github_callback_requires_code(monkeypatch):
+    monkeypatch.setattr(
+        "app.routes.auth.exchange_code_for_token",
+        lambda code: {"access_token": "fake-token"},
+    )
+    monkeypatch.setattr(
+        "app.routes.auth.create_session",
+        lambda access_token: "signed-session",
+    )
+
+    response = client.get("/auth/github/callback")
+
+    assert response.status_code == 400
+    assert response.json()["detail"] == "Missing GitHub OAuth code"
+
+
+def test_analyze_commits_includes_history_and_authors():
+    commits = [
+        {
+            "sha": "abc123",
+            "author": {"login": "alice"},
+            "commit": {
+                "message": "Fix login flow\n\nDetails",
+                "author": {"name": "Alice", "date": "2024-01-02T00:00:00Z"}
+            },
+            "html_url": "https://example.com/commit/abc123"
+        },
+        {
+            "sha": "def456",
+            "author": {"login": "bob"},
+            "commit": {
+                "message": "Add dashboard stats",
+                "author": {"name": "Bob", "date": "2024-01-01T00:00:00Z"}
+            },
+            "html_url": "https://example.com/commit/def456"
+        }
+    ]
+
+    analysis = analyze_commits(commits)
+
+    assert analysis["total_analyzed"] == 2
+    assert analysis["history"][0]["author"] == "alice"
+    assert analysis["history"][0]["message"] == "Fix login flow"
+    assert analysis["history"][1]["author"] == "bob"
+
+
+def test_system_prompt_requires_repository_evidence():
+    assert "Repository descriptions are weak evidence" in SYSTEM_PROMPT
+    assert "source file analysis" in SYSTEM_PROMPT
+    assert "retrieved code snippets" in SYSTEM_PROMPT

frontend/src/pages/Repository.jsx

@@ -98,8 +98,9 @@
                 <strong>{analysis.architecture.backend || "Source"}</strong>
               </div>
               <p className="muted">
-                {analysis.architecture.analyzed_source_files || 0} source files ·{" "}
-                {analysis.architecture.module_relationships || 0} detected module relationships
+                {analysis.architecture.analyzed_source_files || 0} source files
+                · {analysis.architecture.module_relationships || 0} detected
+                module relationships
               </p>
               {analysis.architecture.entry_points?.length > 0 && (
                 <div className="source-list">
@@ -113,12 +114,14 @@
               )}
               {analysis.architecture.module_graph?.length > 0 && (
                 <div className="source-list">
-                  {analysis.architecture.module_graph.slice(0, 12).map((edge) => (
-                    <div key={`${edge.from}-${edge.to}-${edge.import}`}>
-                      <code>{edge.from}</code>
-                      <span>→ {edge.to}</span>
-                    </div>
-                  ))}
+                  {analysis.architecture.module_graph
+                    .slice(0, 12)
+                    .map((edge) => (
+                      <div key={`${edge.from}-${edge.to}-${edge.import}`}>
+                        <code>{edge.from}</code>
+                        <span>→ {edge.to}</span>
+                      </div>
+                    ))}
                 </div>
               )}
             </article>
@@ -147,6 +150,22 @@
                   : "no setup detected"}
               </p>
             </article>
+            <article className="panel wide-panel">
+              <p className="eyebrow">Commit history</p>
+              <h2>{analysis.commits.total_analyzed} total commits</h2>
+              <div className="source-list">
+                {(analysis.commits.history || []).slice(0, 8).map((commit) => (
+                  <div key={commit.sha || `${commit.author}-${commit.date}`}>
+                    <code>{commit.message}</code>
+                    <span>
+                      {commit.author} ·{" "}
+                      {new Date(commit.date).toLocaleDateString()}
+                      {commit.date ? "" : ""}
+                    </span>
+                  </div>
+                ))}
+              </div>
+            </article>
             <article className="panel wide-panel">
               <p className="eyebrow">Important files</p>
               <h2>