How primeintellect-ai/lab-cookbook works
Lab Cookbook is a workspace of learning guides, runnable configurations, and local agent environments for Prime Intellect Lab. Its principal user journey is to follow a guide, select an environment and configuration, then evaluate or train an agent; prompt optimization and deployment are additional documented capabilities. The graph includes representative environment modules and specialized search, coding, and multimodal examples. README statements support the workspace and capability links; sampled source confirms selected tool and task implementations, but not a single shared runtime orchestrator.
Main entry points: guides Cookbook Guides · configs Eval Configs · calendar_problem.py Calendar Problem · gepa GEPA Configs
Components 26
Learning Guides
- Cookbook Guides
guides
Run Configurations
- Eval Configs
configs - GEPA Configs
configs/gepa - RL Configs
configs/rl
Agent Environments
- Reverse Text
environments/reverse_text/reverse_text.py - Wordle
environments/wordle/wordle.py - Shape Detective
environments/shape_detective/shape_detective.py - Calendar Problem
environments/calendar_scheduling/calendar_problem.py - Calendar Tools
environments/calendar_scheduling/calendar_scheduling.py - Calendar TUI
environments/calendar_scheduling/calendar_scheduling_tui.py - Python Math
environments/math_python/math_python.py - GSM8K
environments/gsm8k/gsm8k.py - Ethics Debate
environments/ethics_debate/ethics_debate.py - Simple Judge
environments/simple_judge/simple_judge.py
Search Environments
- Wiki Search
environments/wiki_search/wiki_search.py - Patent Q&A
environments/patent_search/basic_patent_q_and_a/basic_patent_q_and_a.py - Advanced Patent Q&A
environments/patent_search/advanced_patent_q_and_a/advanced_patent_q_and_a.py - Patent Analysis
environments/patent_search/patent_technical_analysis/patent_technical_analysis.py - Wikispeedia Agent
environments/langchain_deep_agents_wikispeedia/langchain_deep_agents_wikispeedia.py - Wiki Graph
environments/langchain_deep_agents_wikispeedia/wiki_graph.py - Deep Agents
environments/langchain_deep_agents/langchain_deep_agents.py
Specialized Environments
- Code Search
environments/swe_grep/swe_grep.py - Coding Agent
environments/opencode_harbor/opencode_harbor.py
Other components
- Lab User
- Agent
- Prime Intellect Lab
Connections 13
- Lab User → Cookbook Guides: follows
README.md - Cookbook Guides → Eval Configs: uses recipes
README.md - Cookbook Guides → Reverse Text: builds environments
README.md - Cookbook Guides → Agent: teaches evaluating
README.md - Cookbook Guides → Agent: teaches training
README.md - Cookbook Guides → Agent: teaches optimizing
README.md - Cookbook Guides → Prime Intellect Lab: covers deployment
README.md - Eval Configs → Agent: evaluates
README.md - GEPA Configs → Agent: optimizes prompts
README.md - RL Configs → Agent: trains
README.md - Calendar Tools → Calendar Problem: uses task model
environments/calendar_scheduling/calendar_scheduling.py - Calendar TUI → Calendar Problem: renders task
environments/calendar_scheduling/calendar_scheduling_tui.py - Wikispeedia Agent → Wiki Graph: loads graph
environments/langchain_deep_agents_wikispeedia/langchain_deep_agents_wikispeedia.py