feat: deep-research-agent — LangGraph + BMasterAI telemetry by ellucas-creator · Pull Request #50 · travis-burmaster/bmasterai

ellucas-creator · 2026-03-14T21:51:41Z

Deep Research Agent

A multi-step web research agent built with LangGraph and instrumented with BMasterAI logging and telemetry. Inspired by langchain-ai/deepagents.

Architecture

planner → web_searcher → analyzer → reflector
                                         ↓
                             needs_more_research?
                              /               \
                            yes (≤2x)         no
                             ↓                ↓
                        web_searcher      synthesizer → END
                       (follow-ups)

BMasterAI Integration

Feature	Where used
`configure_logging`	Startup
`monitor.track_agent_start/stop`	Every node
`monitor.track_llm_call`	Every Claude API call
`monitor.track_task_duration`	Per-node timing
`bm.log_event(EventType.LLM_CALL)`	Before each LLM call
`bm.log_event(EventType.TOOL_USE)`	Each Tavily search
`bm.log_event(EventType.DECISION_POINT)`	Planner + reflector routing
`bm.log_reasoning_chain`	Planner sub-question breakdown
`monitor.get_agent_dashboard()`	Final telemetry printout
`logs/research.jsonl`	Structured JSON for analytics

Usage

pip install -r requirements.txt
cp .env.example .env  # add ANTHROPIC_API_KEY + TAVILY_API_KEY
python main.py "What is the current state of AI agents in enterprise software?"

… by deepagents) - Planner decomposes research topic into 3-5 sub-questions with reasoning chain logging - Web Searcher runs Tavily searches per sub-question, logs each tool use - Analyzer synthesizes results into findings, logs LLM calls + token usage - Reflector evaluates quality (1-10 score), conditionally loops for more research (max 2x) - Synthesizer produces structured report (Executive Summary, Key Findings, Analysis, Limitations) - Full BMasterAI instrumentation: agent lifecycle, LLM calls, tool use, decision points, reasoning chains - Telemetry written to logs/research.jsonl for downstream analytics - Inspired by langchain-ai/deepagents deep_research example

claude · 2026-03-14T21:51:56Z

Claude finished @ellucas-creator's task —— View job

Code Review: Deep Research Agent

I've completed a comprehensive review of this LangGraph + BMasterAI integration. Here's my detailed feedback:

Review Checklist

✅ Strengths

Architecture & Design

Excellent LangGraph workflow design with clear separation of concerns
Well-structured state management using TypedDict for type safety
Clean conditional routing logic for reflection loops with max iteration limit
Comprehensive BMasterAI telemetry integration throughout the pipeline

Code Quality

Good docstrings and inline documentation
Consistent error handling patterns across agents
Proper environment variable validation with clear error messages
Clean JSON parsing with graceful fallbacks

Security

API keys properly handled via environment variables
No hardcoded secrets in code
Appropriate .gitignore excluding sensitive files

🔍 Issues & Recommendations

High Priority

Missing Test Coverage - agents.py:1-386
- No unit tests for any agent functions
- No integration tests for the LangGraph pipeline
- No tests for error handling scenarios
- Recommendation: Add test files covering agent logic, API failures, and edge cases

Insufficient Input Validation - main.py:92-99

# Current: basic strip() only
topic = input("Research question: ").strip()
if not topic:
    topic = "What is the current state of multi-agent AI systems in 2026?"

No validation for topic length or content
Recommendation: Add length limits and content validation

Resource Leak Risk - agents.py:28-29
```
monitor = get_monitor()
monitor.start_monitoring()
```
- Monitor started but never explicitly stopped
- Recommendation: Add proper cleanup in main.py or use context manager

Medium Priority

Error Information Loss - agents.py:110-112
```
except Exception as e:
    sub_questions = [state["topic"]]
    reasoning = f"Parsing failed ({e}), falling back to direct search"
```
- Exception details lost in user-facing output
- Recommendation: Log full exception details via BMasterAI for debugging
Inefficient API Usage - agents.py:162-186
```
for q in questions:
    # Sequential API calls
    result = tavily.search(query=q, ...)
```
- Sequential Tavily calls could be parallelized
- Recommendation: Use asyncio or concurrent.futures for parallel searches
Memory Growth - agents.py:158-160
```
all_results = list(state.get("search_results", []))
all_sources = list(state.get("sources", []))
```
- Results accumulate across reflection loops without cleanup
- Recommendation: Consider result size limits or pruning strategy

Low Priority

Magic Numbers - agents.py:38
```
MAX_REFLECTION_LOOPS = 2
```
- Hardcoded configuration values
- Recommendation: Move to environment variables or config file
Type Safety Gap - agents.py:57
```
tokens_used=getattr(response, "usage_metadata", {}).get("total_tokens", 0) if hasattr(response, "usage_metadata") else 0
```
- Complex conditional logic for token extraction
- Recommendation: Create helper function with proper typing

⚡ Performance Considerations

API Rate Limiting: No rate limiting for Tavily or Anthropic APIs
Memory Usage: Large search results accumulate in state without bounds
Latency: Sequential operations could benefit from parallelization
Token Efficiency: No token usage optimization for LLM calls

🛡️ Security Assessment

✅ Good Practices

Environment variables for API keys
No sensitive data in version control
Proper .gitignore configuration

⚠️ Areas for Improvement

No input sanitization for research topics
BMasterAI logs might contain sensitive research queries
No rate limiting protection against abuse

📚 Documentation Quality

Excellent

Comprehensive README with usage examples
Clear architecture diagrams
Well-documented BMasterAI integration table
Good inline documentation

Minor Improvements

Add API rate limits to documentation
Include troubleshooting section
Document environment variable options

🚀 Recommended Next Steps

Add comprehensive test suite (unit + integration)
Implement proper resource cleanup in main.py
Add input validation and sanitization
Consider parallelizing Tavily searches for performance
Add configuration management for magic numbers
Implement error logging improvements

📊 Overall Assessment

This is a well-architected example that demonstrates excellent LangGraph usage and comprehensive BMasterAI integration. The code follows good Python practices with clear separation of concerns. However, it lacks test coverage and could benefit from better error handling and performance optimizations.

Recommendation: ✅ Approve with suggested improvements - This is ready for merge as an example, but consider the test coverage and input validation improvements for production use.

claude · 2026-03-14T22:14:54Z

Claude finished @ellucas-creator's task —— View job

Code Review: Deep Research Agent

I've completed a comprehensive review of this LangGraph + BMasterAI integration. Here's my detailed feedback:

Review Checklist

✅ Strengths

Architecture & Design

Excellent LangGraph workflow design with clear separation of concerns
Well-structured state management using TypedDict for type safety
Clean conditional routing logic for reflection loops with max iteration limit
Comprehensive BMasterAI telemetry integration throughout the pipeline

Code Quality

Good docstrings and inline documentation
Consistent error handling patterns across agents
Proper environment variable validation with clear error messages
Clean JSON parsing with graceful fallbacks

Security

API keys properly handled via environment variables
No hardcoded secrets in code
Appropriate .gitignore excluding sensitive files

🔍 Issues & Recommendations

High Priority

Missing Test Coverage - agents.py:1-386
- No unit tests for any agent functions
- No integration tests for the LangGraph pipeline
- No tests for error handling scenarios
- Recommendation: Add test files covering agent logic, API failures, and edge cases

Insufficient Input Validation - main.py:92-99

# Current: basic strip() only
topic = input("Research question: ").strip()
if not topic:
    topic = "What is the current state of multi-agent AI systems in 2026?"

No validation for topic length or content
Recommendation: Add length limits and content validation

Resource Leak Risk - agents.py:28-29
```
monitor = get_monitor()
monitor.start_monitoring()
```
- Monitor started but never explicitly stopped
- Recommendation: Add proper cleanup in main.py or use context manager

Medium Priority

Error Information Loss - agents.py:110-112
```
except Exception as e:
    sub_questions = [state["topic"]]
    reasoning = f"Parsing failed ({e}), falling back to direct search"
```
- Exception details lost in user-facing output
- Recommendation: Log full exception details via BMasterAI for debugging
Inefficient API Usage - agents.py:162-186
```
for q in questions:
    # Sequential API calls
    result = tavily.search(query=q, ...)
```
- Sequential Tavily calls could be parallelized
- Recommendation: Use asyncio or concurrent.futures for parallel searches
Memory Growth - agents.py:158-160
```
all_results = list(state.get("search_results", []))
all_sources = list(state.get("sources", []))
```
- Results accumulate across reflection loops without cleanup
- Recommendation: Consider result size limits or pruning strategy

Low Priority

Magic Numbers - agents.py:38
```
MAX_REFLECTION_LOOPS = 2
```
- Hardcoded configuration values
- Recommendation: Move to environment variables or config file
Type Safety Gap - agents.py:57
```
tokens_used=getattr(response, "usage_metadata", {}).get("total_tokens", 0) if hasattr(response, "usage_metadata") else 0
```
- Complex conditional logic for token extraction
- Recommendation: Create helper function with proper typing

⚡ Performance Considerations

API Rate Limiting: No rate limiting for Tavily or Anthropic APIs
Memory Usage: Large search results accumulate in state without bounds
Latency: Sequential operations could benefit from parallelization
Token Efficiency: No token usage optimization for LLM calls

🛡️ Security Assessment

✅ Good Practices

Environment variables for API keys
No sensitive data in version control
Proper .gitignore configuration

⚠️ Areas for Improvement

No input sanitization for research topics
BMasterAI logs might contain sensitive research queries
No rate limiting protection against abuse

📚 Documentation Quality

Excellent

Comprehensive README with usage examples
Clear architecture diagrams
Well-documented BMasterAI integration table
Good inline documentation

Minor Improvements

Add API rate limits to documentation
Include troubleshooting section
Document environment variable options

🚀 Recommended Next Steps

Add comprehensive test suite (unit + integration)
Implement proper resource cleanup in main.py
Add input validation and sanitization
Consider parallelizing Tavily searches for performance
Add configuration management for magic numbers
Implement error logging improvements

📊 Overall Assessment

This is a well-architected example that demonstrates excellent LangGraph usage and comprehensive BMasterAI integration. The code follows good Python practices with clear separation of concerns. However, it lacks test coverage and could benefit from better error handling and performance optimizations.

Recommendation: ✅ Approve with suggested improvements - This is ready for merge as an example, but consider the test coverage and input validation improvements for production use.

docs: add deep-research-agent and langgraph-viral-youtube to root README

6efc184

travis-burmaster merged commit b02369e into main Mar 14, 2026
1 check passed

travis-burmaster deleted the feat/deep-research-agent branch March 14, 2026 22:15

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: deep-research-agent — LangGraph + BMasterAI telemetry#50

feat: deep-research-agent — LangGraph + BMasterAI telemetry#50
travis-burmaster merged 2 commits intomainfrom
feat/deep-research-agent

ellucas-creator commented Mar 14, 2026

Uh oh!

claude bot commented Mar 14, 2026 •

edited

Loading

Uh oh!

claude bot commented Mar 14, 2026 •

edited

Loading

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

ellucas-creator commented Mar 14, 2026

Deep Research Agent

Architecture

BMasterAI Integration

Usage

Uh oh!

claude bot commented Mar 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Code Review: Deep Research Agent

Review Checklist

✅ Strengths

🔍 Issues & Recommendations

High Priority

Medium Priority

Low Priority

⚡ Performance Considerations

🛡️ Security Assessment

📚 Documentation Quality

🚀 Recommended Next Steps

📊 Overall Assessment

Uh oh!

claude bot commented Mar 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Code Review: Deep Research Agent

Review Checklist

✅ Strengths

🔍 Issues & Recommendations

High Priority

Medium Priority

Low Priority

⚡ Performance Considerations

🛡️ Security Assessment

📚 Documentation Quality

🚀 Recommended Next Steps

📊 Overall Assessment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

claude bot commented Mar 14, 2026 •

edited

Loading

claude bot commented Mar 14, 2026 •

edited

Loading