Your AI-powered Fantasy Premier League co-manager that provides data-driven insights, transfer suggestions, and tactical advice.
FPL Gaffer is an intelligent agentic AI system designed to assist Fantasy Premier League (FPL) managers with informed decision-making. Built using LangGraph, it combines real-time data retrieval, intelligent tool orchestration, and response validation to provide accurate, hallucination-free advice for your FPL team.
Unlike generic chatbots, FPL Gaffer:
- 🎯 Never hallucinates - Every suggestion is backed by real data
- 🔄 Self-validates - Automatically checks responses for accuracy before sending
- 🧠 Learns from mistakes - Validation feedback loops back to gather missing data
- 💬 Speaks FPL - Uses community language and understands the game deeply
- 📱 WhatsApp-ready - Conversational, plain-text responses perfect for mobile
- Context-Aware Processing: Automatically injects user context (team value, budget, rank) into every interaction
- Smart Tool Selection: LLM-powered analysis determines which data sources to query
- Multi-Tool Orchestration: Seamlessly combines multiple data sources for comprehensive answers
- Validation Loop: Self-corrects by detecting hallucinations and gathering missing data
- Conversation Memory: MySQL checkpointer maintains conversation history for contextual responses
- 📊 Team Analysis: Detailed insights into your current squad, transfers, and finances
- 🔍 Player Research: Stats, form, injuries, and pricing for informed transfers
- 📅 Fixture Analysis: Upcoming schedule planning for captain picks and transfers
- 💰 Budget Planning: Smart suggestions within your exact financial constraints
- 📰 News Integration: Real-time injury updates, press conferences, and expert opinions
- 🎲 Transfer Suggestions: Position-based player recommendations with filters
class WorkflowState(MessagesState):
"""State for the FPL Gaffer workflow."""
# User & gameweek context
user_id: int
user_data: Dict[str, Any]
gameweek_data: Dict
# Tool execution
tool_calls: List[Dict[str, Any]]
tool_results: Dict[str, Any]
# Response generation
response: str
# Validation and control flow
is_retry: bool
retry_count: int
validation_passed: bool
validation_errors: List[str]
validation_suggestions: List[str]The tool selection node uses LLM reasoning to determine which tools are needed:
- Initial requests: Analyzes user query directly
- Retry requests: Processes validation feedback to gather missing data
# Example validation feedback loop:
Validation: "Response mentions 'good fixtures' but no fixture data retrieved"
↓
Retry Tool Selection: Calls get_fixtures_for_range_tool()
↓
New Response: Based on actual fixture dataPrevents hallucinations by checking:
- ✅ All player suggestions exist in tool results
- ✅ Price recommendations match user budget
- ✅ Fixture claims backed by actual data
- ✅ Statistics match retrieved information
- ✅ All user questions addressed with real data
| Tool | Description | Use Case |
|---|---|---|
news_search_tool |
Search FPL news, injuries, press conferences | Player news, expert opinions, updates |
get_user_team_info_tool |
Retrieve user's squad, transfers, finances | Team analysis, budget planning |
get_players_by_position_tool |
Filter players by position & max price | Transfer targets, squad building |
get_player_data_tool |
Get detailed player stats, form, injuries | Player comparison, transfer decisions |
get_fixtures_for_range_tool |
Get upcoming fixtures for X gameweeks | Captain picks, fixture planning |
User: "I have £1.5m in the bank. Any good midfielders I should look at?"
FPL Gaffer: "Nice budget to work with! Let me check the mid-priced midfielders
for you... Palmer (£6.8m) is looking cracking right now - 3 goals in last 4 games
and Chelsea's fixtures are solid. Saka (£9.0m) is another shout if you can stretch
the budget a bit. Both on penalties too which is a bonus!"
User: "What are the fixtures like for the next 3 gameweeks?"
FPL Gaffer: "Right, let's look ahead...GW10-12 is massive for City assets - they've
got Bournemouth (H), Brighton (A), and Spurs (H). Haaland differential could be
huge here. Arsenal have a mixed bag with Liverpool (A) sandwiched between two
decent fixtures. Villa's got the best run though - all green on the ticker!"
User: "How's my team looking?"
FPL Gaffer: "Your team's solid mate! Sitting at £102.5m value with £0.5m ITB.
You've got the template sorted with Salah and Haaland up top. Defense might need
a look though - you've got three players with tough fixtures coming up. Fancy
making a move there?"
- Agentic AI Design: Not just a chatbot - FPL Gaffer reasons about what data it needs
- Self-Healing Responses: Validation loops ensure accuracy before delivery
- Modular Architecture: Clean separation of concerns for easy maintenance
- Production-Ready: MySQL persistence, error handling, retry logic
- Domain Expertise: Understands FPL strategy, not just player stats
- Community Language: Speaks like an FPL veteran
- Budget-Aware: Never suggests transfers you can't afford
- Fixture-Conscious: Considers short and long-term planning
- No Hallucinations: Only suggests players and strategies backed by data
- Conversational: Feels like chatting with your FPL-obsessed mate
- Mobile-First: Designed for WhatsApp, perfect for deadline decisions
- Memory: Remembers your previous conversations and team decisions
This project is licensed under the MIT License - see the LICENSE file for details.
DejusDevspace
- GitHub: @DejusDevsapce
- LinkedIn: Ojomideju Adejo
- Twitter: @d3ju.ai
- LangGraph for the agent framework
- Fantasy Premier League API for FPL data
- The FPL community for inspiration and insights
⚽ Built with passion for FPL managers ⚽
If you find this project helpful, please consider giving it a ⭐!