# Index

| File | Purpose | Subsystem | Symbols | Used by |
|------|---------|-----------|---------|---------|
| `app.py` | Drop-in entry point that demonstrates how to use the topogpt3 package. | root | 5 | 0 |
| `eval/analyze.py` | Aggregate HumanEval result JSONL files into a summary table. | eval | 5 | 0 |
| `eval/analyze_results.py` | Analyze a HumanEval JSONL produced by harness.py. | eval | 4 | 0 |
| `eval/diag_static.py` | Diagnostico estatico de un checkpoint TopoGPT3 congelado. | eval | 5 | 0 |
| `eval/governor.py` | Streaming + governance for autoregressive generation. | eval | 20 | 1 |
| `eval/governor_smoke.py` | Smoke test for eval.governor (TokenStream + GenerationGovernor). | eval | 7 | 0 |
| `eval/harness.py` | Harness for evaluating TopoGPT3 on HumanEval (164 problems). | eval | 12 | 3 |
| `eval/hodge_cm_ablation.py` | Hodge-CM offline ablation on real TopoGPT3 checkpoint. | eval | 6 | 0 |
| `eval/integration_smoke.py` | End-to-end smoke test of all P0+P1 components working together. | eval | 1 | 0 |
| `eval/noise_analysis.py` | Analisis post-hoc del noise sweep. | eval | 3 | 0 |
| `eval/noise_sweep.py` | Barrido de ruido en los pesos espectrales del checkpoint TopoGPT3. | eval | 4 | 1 |
| `eval/repair.py` | Self-repair loop on top of a greedy JSONL. | eval | 6 | 0 |
| `eval/report.py` | Aggregate every JSONL in eval/runs into a final report. | eval | 6 | 0 |
| `eval/samplers.py` | Registry of sampler constructors for the HumanEval harness. | eval | 7 | 1 |
| `eval/sandbox.py` | Sandbox for executing model-generated code during HumanEval evaluation. | eval | 9 | 2 |
| `eval/sandbox_smoke.py` | Smoke test for eval.sandbox. | eval | 1 | 0 |
| `eval/smoke.py` | Smoke test: load the TopoGPT3 checkpoint and produce a small completion. | eval | 2 | 0 |
| `eval/temp_sweep.py` | Barrido de temperatura x top-k sobre HumanEval. | eval | 5 | 0 |
| `infer_exploitgym.py` | Standalone inference script for TopoExploit. | root | 11 | 0 |
| `install.sh` | - | root | 0 | 0 |
| `run_exploitgym.sh` | TopoExploit: 125M params trained on ExploitGym  Usage: ./run_exploitgym.sh prepare     -- clone... | root | 0 | 0 |
| `run_exploitgym_v2.sh` | - | root | 4 | 0 |
| `run_hodge_cm_ablation.sh` | Hodge-CM offline ablation runner (GPU si el build torch lo permite, si no CPU). | root | 0 | 0 |
| `run_merged.sh` | - | root | 7 | 0 |
| `synthetic_dataset.py` | Synthetic Dataset Generator for TopoGPT2. | root | 35 | 1 |
| `tests/test_jlens.py` | TestValidPositionMask: Feature: valid_position_mask excludes attention-sink and final positions. | tests | 51 | 0 |
| `tests/test_lens_model.py` | TestTopoGPT3LensConfig: Feature: TopoGPT3LensConfig provides centralized adapter configuration. | tests | 34 | 0 |
| `topogpt3/__init__.py` | TopoGPT3: complex-valued spectral language model for code. | topogpt3 | 0 | 8 |
| `topogpt3/__main__.py` | main: TopoGPT3 entry point. | topogpt3 | 1 | 0 |
| `topogpt3/api_server.py` | OpenAI-compatible HTTP API server so TopoGPT3 can be used as a backend for coding agents (e.g. | topogpt3 | 46 | 1 |
| `topogpt3/continuation.py` | Auto-continuation engine: detects truncated responses and feeds the last incomplete lines back... | topogpt3 | 5 | 3 |
| `topogpt3/ewc.py` | Elastic Weight Consolidation (EWC) and Experience Replay for TopoGPT3. | topogpt3 | 11 | 1 |
| `topogpt3/exploitgym_config.py` | TopoExploit Training Configuration  122M parameter model trained on ExploitGym... | topogpt3 | 3 | 1 |
| `topogpt3/exploitgym_loader.py` | ExploitGym Data Loader for TopoExploit  Clones the ExploitGym repo and tokenizes task data into... | topogpt3 | 19 | 3 |
| `topogpt3/hodge_cm.py` | Fases 2-5: martillo Hodge-CM para TopoGPT3 / modelo espectral continuo. | topogpt3 | 15 | 1 |
| `topogpt3/inference.py` | TopoGPT3 inference engine. | topogpt3 | 54 | 2 |
| `topogpt3/inference_hrm.py` | TopoGPT3.1: Hierarchical Recursive Reasoning Inference Engine. | topogpt3 | 76 | 2 |
| `topogpt3/jlens.py` | TopoGPT3JLensFitConfig: Centralized configuration for Jacobian lens fitting. | topogpt3 | 29 | 4 |
| `topogpt3/lens_model.py` | LensModel: What the lens needs from a model. | topogpt3 | 29 | 5 |
| `topogpt3/merged_config.py` | TopoMerged: 6-tier curriculum merging TopoGPT3 code tasks with ExploitGym exploit tasks. | topogpt3 | 6 | 2 |
| `topogpt3/model.py` | TopoGPT2: Quaternion-Enhanced Topological Transformer Language Model  Author: Gris Iscomeback... | topogpt3 | 187 | 13 |
| `topogpt3/train.py` | TopoGPT3: Grassmannian / Berry-Holonomy extension of TopoGPT2  Author: Gris Iscomeback License... | topogpt3 | 64 | 4 |
| `topogpt3/vanilla_control.py` | Vanilla transformer control for TopoGPT3 Hodge-CM ablation. | topogpt3 | 26 | 0 |
| `transfer_weights.py` | Transfer weights from a smaller TopoGPT model to a larger one. | root | 3 | 0 |