feat: convert to async/await with parallel agent execution via asyncio.gather

app/agents/bug_detection_agent.py

@@ -7,7 +7,7 @@ from app.core.prompts import BUG_DETECTION_PROMPT
 from app.models.issue import IssueReport, Bug, Warning, Suggestion
 
 
-def run(local_path: str) -> IssueReport:
+async def run(local_path: str) -> IssueReport:
     """
     Detect bugs, warnings and suggestions in a repository.
     """
@@ -53,7 +53,7 @@ Code Samples:
 
     # Step 5 - Call LLM
     prompt = BUG_DETECTION_PROMPT.format(code_info=code_info)
-    llm_response = call_llm(prompt)
+    llm_response = await call_llm(prompt)
 
     # Step 6 - Parse response
     try:

app/agents/code_review_agent.py

@@ -6,7 +6,7 @@ from app.models.review import ReviewSuggestions
 from app.models.repository import RepositoryMetadata
 
 
-def run(local_path: str, metadata: RepositoryMetadata) -> ReviewSuggestions:
+async def run(local_path: str, metadata: RepositoryMetadata) -> ReviewSuggestions:
     """
     Perform AI-powered code review on repository.
     """
@@ -36,7 +36,7 @@ def run(local_path: str, metadata: RepositoryMetadata) -> ReviewSuggestions:
     # Step 3 - Call LLM
     source_code = "\n\n".join(source_samples)
     prompt = CODE_REVIEW_PROMPT.format(source_code=source_code)
-    llm_response = call_llm(prompt)
+    llm_response = await call_llm(prompt)
 
     # Step 4 - Parse response
     try:

app/agents/repo_analysis_agent.py

@@ -6,7 +6,7 @@ from app.core.prompts import REPO_ANALYSIS_PROMPT
 from app.models.repository import RepositoryMetadata
 
 
-def run(repo_url: str, local_path: str) -> RepositoryMetadata:
+async def run(repo_url: str, local_path: str) -> RepositoryMetadata:
     """
     Analyze a repository and return structured metadata.
     """
@@ -32,7 +32,7 @@ Functions Found: {sum(len(r['functions']) for r in ast_results)}
 
     # Step 3 - Call LLM
     prompt = REPO_ANALYSIS_PROMPT.format(repo_info=repo_info)
-    llm_response = call_llm(prompt)
+    llm_response = await call_llm(prompt)
 
     # Step 4 - Parse response
     try:

app/agents/report_generator_agent.py

@@ -7,7 +7,7 @@ from app.models.review import ReviewSuggestions, GeneratedTests
 from app.models.report import EngineeringReport
 
 
-def run(
+async def run(
     metadata: RepositoryMetadata,
     issues: IssueReport,
     tests: GeneratedTests,
@@ -56,7 +56,7 @@ Summary: {review.summary}
         tests=tests_summary,
         review=review_summary
     )
-    full_report = call_llm(prompt, max_tokens=3000)
+    full_report = await call_llm(prompt, max_tokens=3000)
 
     # Step 3 - Build final report
     report = EngineeringReport(

app/agents/test_generation_agent.py

@@ -5,7 +5,7 @@ from app.core.prompts import TEST_GENERATION_PROMPT
 from app.models.review import GeneratedTests
 
 
-def run(local_path: str) -> GeneratedTests:
+async def run(local_path: str) -> GeneratedTests:
     """
     Generate pytest test cases for uncovered functions.
     """
@@ -48,7 +48,7 @@ def run(local_path: str) -> GeneratedTests:
     # Step 3 - Call LLM
     source_code = "\n\n".join(source_code_samples)
     prompt = TEST_GENERATION_PROMPT.format(source_code=source_code)
-    llm_response = call_llm(prompt, max_tokens=3000)
+    llm_response = await call_llm(prompt, max_tokens=3000)
 
     # Step 4 - Count generated tests
     test_count = llm_response.count("def test_")

app/api/routes.py

@@ -1,3 +1,4 @@
+import asyncio
 from fastapi import APIRouter, HTTPException
 from app.models.repository import RepositoryRequest, RepositoryResponse
 from app.models.report import EngineeringReport
@@ -19,7 +20,7 @@ def health_check():
 
 
 @router.post("/analyze", response_model=EngineeringReport)
-def analyze_repository(request: RepositoryRequest):
+async def analyze_repository(request: RepositoryRequest):
     """
     Full pipeline:
     1. Clone repository
@@ -33,11 +34,15 @@ def analyze_repository(request: RepositoryRequest):
         local_path = clone_repository(request.github_url)
 
         # Step 2 - Run agents in order
-        metadata = repo_analysis_agent.run(request.github_url, local_path)
-        issues = bug_detection_agent.run(local_path)
-        tests = test_generation_agent.run(local_path)
-        review = code_review_agent.run(local_path, metadata)
-        report = report_generator_agent.run(metadata, issues, tests, review)
+        metadata = await repo_analysis_agent.run(request.github_url, local_path)
+
+        issues, tests, review = await asyncio.gather(
+            bug_detection_agent.run(local_path),
+            test_generation_agent.run(local_path),
+            code_review_agent.run(local_path, metadata),
+        )
+
+        report = await report_generator_agent.run(metadata, issues, tests, review)
 
         return report
 
@@ -51,7 +56,7 @@ def analyze_repository(request: RepositoryRequest):
 
 
 @router.post("/quick-analyze", response_model=RepositoryResponse)
-def quick_analyze(request: RepositoryRequest):
+async def quick_analyze(request: RepositoryRequest):
     """
     Only run repository analysis agent.
     Faster, for testing purposes.
@@ -60,7 +65,7 @@ def quick_analyze(request: RepositoryRequest):
 
     try:
         local_path = clone_repository(request.github_url)
-        metadata = repo_analysis_agent.run(request.github_url, local_path)
+        metadata = await repo_analysis_agent.run(request.github_url, local_path)
 
         return RepositoryResponse(
             success=True,

app/core/llm.py

@@ -6,7 +6,7 @@ settings = get_settings()
 OPENROUTER_API_URL = "https://openrouter.ai/api/v1/chat/completions"
 
 
-def call_llm(prompt: str, system_prompt: str = None, max_tokens: int = 2000) -> str:
+async def call_llm(prompt: str, system_prompt: str = None, max_tokens: int = 2000) -> str:
     """
     Call OpenRouter API with any open source model.
     Returns the response text.