How vox304/k5-l3a-day4-vocongminh-2a202603520-reactloopagent works
This repository is a Python lab for building a status-investigation agent, not a finished agent product: agent.py currently raises NotImplementedError. The intended user workflow is to submit a question through the CLI, have an LLM request status tools, and receive tool results in the conversation before a final answer. The lab also supports snapshot/live status data, an optional multi-provider bonus toolset, and evaluation with an LLM judge. The agent-loop relationships are documented, but its runtime implementation is not present in the sampled source.
Main entry points: run_evals.py Eval runner · providers.py LLM adapter · run.py CLI runner · agent.py Agent loop
Components 14
Agent runtime
- CLI runner
run.py - Agent loop
agent.py - LLM adapter
providers.py
Status investigation
- Core status tools
status_tools.py - Status snapshots
fixtures
Bonus tools
- Bonus status tools
bonus/tools.py - Bonus snapshots
bonus/fixtures
Evaluation
- Eval runner
evals/run_evals.py - Eval scenarios
evals/scenarios.py - Eval cases
evals/cases.json - Case grader
evals/grading.py
Other components
- User
- LLM service
- Status APIs
Connections 22
- User → CLI runner: submits question
README.md - CLI runner → Agent loop: runs agent
run.py - CLI runner → LLM adapter: configures provider
run.py - CLI runner → Core status tools: selects core tools
run.py - CLI runner → Bonus status tools: selects bonus tools
run.py - Agent loop → LLM adapter: requests completion
README.md - LLM adapter → LLM service: sends requests
providers.py - LLM service → LLM adapter: returns response
providers.py - Agent loop → Core status tools: executes tool calls
README.md - Agent loop → Bonus status tools: executes tool calls
README.md - Core status tools → Status snapshots: reads snapshot
status_tools.py - Core status tools → Status APIs: fetches live status
status_tools.py - Bonus status tools → Bonus snapshots: reads snapshot
bonus/tools.py - Bonus status tools → Status APIs: fetches live status
bonus/tools.py - Eval runner → Agent loop: runs cases
evals/run_evals.py - Eval runner → LLM adapter: creates providers
evals/run_evals.py - Eval runner → Eval scenarios: selects scenario tools
evals/run_evals.py - Eval scenarios → Core status tools: extends core tools
evals/scenarios.py - Eval runner → Bonus status tools: runs bonus tools
evals/run_evals.py - Eval runner → Eval cases: loads cases
evals/run_evals.py - Eval runner → Case grader: grades results
evals/run_evals.py - Case grader → LLM adapter: requests judge score
evals/grading.py