graphify-out/GRAPH_REPORT.md

Graph Report - C:\Users\astra\Desktop\hackon\ani\DebugZero (2026-04-26)

Corpus Check

• 17 files · ~91,715 words
• Verdict: corpus is large enough that graph structure adds value.

Summary

• 157 nodes · 334 edges · 10 communities detected
• Extraction: 73% EXTRACTED · 27% INFERRED · 0% AMBIGUOUS · INFERRED: 91 edges (avg confidence: 0.69)
• Token cost: 0 input · 0 output

Community Hubs (Navigation)

• [[_COMMUNITY_Community 0|Community 0]]
• [[_COMMUNITY_Community 1|Community 1]]
• [[_COMMUNITY_Community 2|Community 2]]
• [[_COMMUNITY_Community 3|Community 3]]
• [[_COMMUNITY_Community 4|Community 4]]
• [[_COMMUNITY_Community 5|Community 5]]
• [[_COMMUNITY_Community 6|Community 6]]
• [[_COMMUNITY_Community 7|Community 7]]
• [[_COMMUNITY_Community 8|Community 8]]
• [[_COMMUNITY_Community 9|Community 9]]

God Nodes (most connected - your core abstractions)

1. run_live_api_probe() - 19 edges
2. DebugzeroObservation - 17 edges
3. DebugzeroEnvironment - 17 edges
4. DebugzeroAction - 15 edges
5. DebugzeroState - 14 edges
6. main() - 13 edges
7. evaluate_proposer_fixed_set() - 12 edges
8. DebugzeroEnv - 11 edges
9. BugInjectorVisitor - 11 edges
10. infer_bug_operator() - 10 edges

Surprising Connections (you probably didn't know these)

• build_bug_bank() --calls--> create_dataset() [INFERRED] C:\Users\astra\Desktop\hackon\ani\DebugZero\bug_bank.py → C:\Users\astra\Desktop\hackon\ani\DebugZero\training\grpo_train.py
• _collect_verified_bugs() --calls--> inject_bug() [INFERRED] C:\Users\astra\Desktop\hackon\ani\DebugZero\bug_bank.py → C:\Users\astra\Desktop\hackon\ani\DebugZero\server\bug_injector.py
• _bug_difficulty_score() --calls--> compute_ast_distance() [INFERRED] C:\Users\astra\Desktop\hackon\ani\DebugZero\bug_bank.py → C:\Users\astra\Desktop\hackon\ani\DebugZero\server\graders.py
• DebugzeroEnv --calls--> make_env() [INFERRED] C:\Users\astra\Desktop\hackon\ani\DebugZero\client.py → C:\Users\astra\Desktop\hackon\ani\DebugZero\inference.py
• DebugzeroEnv --calls--> make_env() [INFERRED] C:\Users\astra\Desktop\hackon\ani\DebugZero\client.py → C:\Users\astra\Desktop\hackon\ani\DebugZero\eval\api_baseline.py

Communities

Community 0 - "Community 0"

Cohesion: 0.12 Nodes (32): sample_proposer_prompt(), is_effectively_unchanged(), build_mixed_role_dataset(), build_tiny_local_model_and_tokenizer(), build_weighted_proposer_rows(), choose_proposer_bug_focus(), completion_to_text(), create_dataset() (+24 more)

Community 1 - "Community 1"

Cohesion: 0.13 Nodes (21): Action, Entry point for direct execution via uv run or python -m. This function e, DebugzeroEnv, Client for the DebugZero Environment. This client maintains a persistent, Convert DebugzeroAction to JSON payload for step message. Args:, Parse server response into StepResult[DebugzeroObservation]. Args:, Parse server response into State object. Args: payload:, Parse server response into State object. Args: payload: (+13 more)

Community 2 - "Community 2"

Cohesion: 0.13 Nodes (12): BugInjectorVisitor, infer_bug_operator(), _infer_condition_negation(), _infer_loop_boundary_shift(), _infer_off_by_one(), _infer_slice_boundary_corruption(), _infer_wrong_builtin(), _infer_wrong_operator() (+4 more)

Community 3 - "Community 3"

Cohesion: 0.16 Nodes (20): compute_ast_distance(), compute_proposer_reward(), compute_solver_reward(), get_solve_rate(), Computes the string similarity distance between the AST dumps of the original, record_solve_result(), reset_reward_history(), build_prompt() (+12 more)

Community 4 - "Community 4"

Cohesion: 0.23 Nodes (12): _bug_difficulty_score(), BugBank, BugSample, build_bug_bank(), _collect_verified_bugs(), _count_nonempty_lines(), validate_seed(), execute_code() (+4 more)

Community 5 - "Community 5"

Cohesion: 0.28 Nodes (12): build_prompt(), call_env_method(), compact_action_string(), extract_env_error(), extract_python_code(), get_model_code(), main(), make_env() (+4 more)

Community 6 - "Community 6"

Cohesion: 0.24 Nodes (1): main()

Community 7 - "Community 7"

Cohesion: 0.83 Nodes (3): sample_solver_prompt(), summarize_failure_output(), _truncate_text()

Community 8 - "Community 8"

Cohesion: 1.0 Nodes (1): Generate ALL publication-quality training plots for DebugZero README. Data sour

Community 9 - "Community 9"

Cohesion: 1.0 Nodes (1): State for the DebugZero environment, extending default state with seed context.

Knowledge Gaps

• 9 isolated node(s): Action for the DebugZero environment representing the Proposer or Solver inputs., Observation from the DebugZero environment following sandbox execution., State for the DebugZero environment, extending default state with seed context., Generate ALL publication-quality training plots for DebugZero README. Data sour, 4 critical checks: - parse succeeds - mutation actually changed code (+4 more) These have ≤1 connection - possible missing edges or undocumented components.
• Thin community Community 8 (2 nodes): generate_all_plots.py, Generate ALL publication-quality training plots for DebugZero README. Data sour Too small to be a meaningful cluster - may be noise or needs more connections extracted.
• Thin community Community 9 (1 nodes): State for the DebugZero environment, extending default state with seed context. Too small to be a meaningful cluster - may be noise or needs more connections extracted.

Suggested Questions

Questions this graph is uniquely positioned to answer:

• Why does run_live_api_probe() connect Community 5 to Community 0, Community 1, Community 2, Community 3, Community 7? High betweenness centrality (0.272) - this node is a cross-community bridge.
• Why does DebugzeroAction connect Community 1 to Community 3, Community 5, Community 6? High betweenness centrality (0.111) - this node is a cross-community bridge.
• Are the 13 inferred relationships involving run_live_api_probe() (e.g. with DebugzeroAction and DebugzeroEnv) actually correct? run_live_api_probe() has 13 INFERRED edges - model-reasoned connections that need verification.
• Are the 14 inferred relationships involving DebugzeroObservation (e.g. with DebugzeroEnv and Client for the DebugZero Environment. This client maintains a persistent) actually correct? DebugzeroObservation has 14 INFERRED edges - model-reasoned connections that need verification.
• Are the 5 inferred relationships involving DebugzeroEnvironment (e.g. with Entry point for direct execution via uv run or python -m. This function e and DebugzeroAction) actually correct? DebugzeroEnvironment has 5 INFERRED edges - model-reasoned connections that need verification.
• Are the 12 inferred relationships involving DebugzeroAction (e.g. with DebugzeroEnv and Client for the DebugZero Environment. This client maintains a persistent) actually correct? DebugzeroAction has 12 INFERRED edges - model-reasoned connections that need verification.
• Are the 11 inferred relationships involving DebugzeroState (e.g. with DebugzeroEnv and Client for the DebugZero Environment. This client maintains a persistent) actually correct? DebugzeroState has 11 INFERRED edges - model-reasoned connections that need verification.