Upload folder using huggingface_hub

README.md

@@ -4,61 +4,484 @@ emoji: 🧪
 colorFrom: blue
 colorTo: green
 sdk: docker
+app_port: 8000
 pinned: false
 base_path: /web
 ---
 
 # DebugZero Environment
 
-*Our environment extends the Absolute Zero paradigm (Zhao et al., NeurIPS 2025) to adversarial bug-fixing self-play. The Proposer's mutation operators are implemented from scratch using Python AST manipulation across 8 operator types. The verifier adapts Mutahunter's execution pipeline.*
+DebugZero is an OpenEnv environment for training code models through adversarial debugging self-play.
 
-DebugZero is a self-play environment where one LLM plays two roles:
-1. **Proposer**: Injects realistic bugs into clean Python functions using AST-level edits.
-2. **Solver**: Fixes the bugs, practicing on the generated adversarial examples.
+One model plays two roles:
 
-This project enables fully self-improving debugging capabilities without human-curated datasets, verified by a strict Python execution sandbox.
+1. **Proposer**: receives a clean Python function and submits a realistic buggy version.
+2. **Solver**: receives the buggy function and submits a repaired version.
 
-## Setup
+The environment executes the submitted code against tests in a constrained Python sandbox and returns structured OpenEnv observations. The training pipeline turns those observations into scalar rewards for GRPO/Unsloth training.
+
+The goal is to teach an LLM a debugging skill that static supervised examples do not capture well: generating plausible failures, diagnosing them, and repairing code based on executable feedback.
+
+## Submission Links
+
+- **Hugging Face Space**: add the final submitted Space URL here before the deadline.
+- **Training notebook**: [`notebooks/train_colab.ipynb`](notebooks/train_colab.ipynb)
+- **OpenEnv manifest**: [`openenv.yaml`](openenv.yaml)
+
+## Why This Environment Matters
+
+Most code training data shows finished solutions. DebugZero instead creates a loop where the model has to reason about failure:
+
+- What kind of bug would a real programmer accidentally introduce?
+- Does the mutated program still parse and run?
+- Does it fail tests for a meaningful reason?
+- Can the solver recover the original intended behavior?
+
+That makes the environment useful for training debugging, program repair, adversarial test thinking, and execution-grounded code reasoning.
+
+## OpenEnv Integration
+
+DebugZero uses the standard OpenEnv client/server pattern.
+
+The manifest is:
+
+```yaml
+spec_version: 1
+name: debugZero
+type: space
+runtime: fastapi
+app: server.app:app
+port: 8000
+workers: 4
+max_concurrent_envs: 100
+```
+
+The FastAPI app is created with OpenEnv's server helper in [`server/app.py`](server/app.py):
+
+```python
+app = create_app(
+    DebugzeroEnvironment,
+    DebugzeroAction,
+    DebugzeroObservation,
+    env_name="debugZero",
+    max_concurrent_envs=int(os.environ.get("MAX_CONCURRENT_ENVS", "100")),
+)
+```
+
+Clients should interact with the environment through [`DebugzeroEnv`](client.py), not by importing server internals. The client serializes `DebugzeroAction` objects, parses OpenEnv `StepResult` payloads, and exposes the normal `reset`, `step`, and `state` flow.
+
+## Episode Flow
+
+Each episode is a two-turn game over one seed function.
+
+### 1. Reset
+
+`reset()` creates a fresh `DebugzeroState`:
+
+- `episode_id`: new UUID
+- `step_count`: `0`
+- `seed_id`: currently `HumanEval/0`
+- `original_code`: clean HumanEval seed implementation
+- `current_code`: initially the same clean code
+- `role_turn`: `proposer`
+
+The reset observation tells the agent that the proposer acts first and provides the clean function.
+
+### 2. Proposer Step
+
+The proposer sends:
+
+```json
+{
+  "role": "proposer",
+  "code": "<complete mutated Python function>"
+}
+```
+
+The environment:
+
+1. Stores the submitted code as `current_code`.
+2. Runs it with the seed tests using `execute_code`.
+3. Returns an observation with:
+   - `role_next = "solver"`
+   - `tests_passed`
+   - `syntax_error`
+   - truncated `execution_result`
+   - `done = false`
+
+A good proposer submission is syntax-valid, safe to execute, close to the original code, and causes tests to fail.
+
+### 3. Solver Step
+
+The solver sends:
+
+```json
+{
+  "role": "solver",
+  "code": "<complete repaired Python function>"
+}
+```
+
+The environment:
+
+1. Stores the submitted repair as `current_code`.
+2. Runs it against the same tests.
+3. Returns an observation with:
+   - `role_next = "end"`
+   - `tests_passed`
+   - `syntax_error`
+   - truncated `execution_result`
+   - `done = true`
+
+A good solver submission passes tests without syntax errors.
+
+## Action, Observation, and State Schemas
+
+The OpenEnv models live in [`models.py`](models.py).
+
+### Action
+
+`DebugzeroAction` extends OpenEnv `Action`:
+
+| Field | Type | Meaning |
+| --- | --- | --- |
+| `role` | `str` | Either `proposer` or `solver`. |
+| `code` | `str` | The complete buggy or repaired Python function. |
+
+### Observation
+
+`DebugzeroObservation` extends OpenEnv `Observation`:
+
+| Field | Type | Meaning |
+| --- | --- | --- |
+| `role_next` | `str` | Which role should act next. |
+| `current_code` | `str` | Current code after reset or step. |
+| `execution_result` | `str` | Captured stdout/stderr summary from sandbox execution. |
+| `tests_passed` | `bool` | Whether the submitted code passed the environment tests. |
+| `syntax_error` | `bool` | Whether parsing or execution produced a syntax error. |
+| `done` | `bool` | OpenEnv completion flag. |
+| `reward` | `float` | Server currently returns `0.0`; training code computes shaped rewards externally. |
+
+### State
+
+`DebugzeroState` extends OpenEnv `State`:
+
+| Field | Type | Meaning |
+| --- | --- | --- |
+| `seed_id` | `str` | Identifier for the seed task. |
+| `original_code` | `str` | Clean reference code. |
+| `current_code` | `str` | Latest proposer or solver code. |
+| `role_turn` | `str` | Internal turn marker: `proposer`, `solver`, or `end`. |
+
+## Reward and Grading Logic
+
+DebugZero separates **verification** from **reward shaping**.
+
+- The OpenEnv server is the verifier. It runs submitted code and returns observations.
+- The training layer is the grader. It reads `tests_passed`, `syntax_error`, plausibility, and solve history, then computes scalar rewards.
+
+This is intentional: the same environment can support different reward rubrics without changing the OpenEnv API.
+
+### Server Verifier
+
+[`DebugzeroEnvironment.step`](server/debugZero_environment.py) always executes code and reports the result, but currently returns `reward=0.0` in the observation. The meaningful reward is computed by the training code from the observation fields.
+
+For proposer actions:
+
+- Syntax error: bad mutation.
+- Tests still pass: mutation did not create a useful bug.
+- Tests fail without syntax error: likely useful bug.
+
+For solver actions:
+
+- Tests pass without syntax error: solved.
+- Tests fail or syntax error: not solved.
+
+### Proposer Reward
+
+Implemented in [`training/rewards.py`](training/rewards.py):
+
+```python
+reward = validity + plausibility + learnability
+```
+
+Components:
+
+| Component | Logic | Reason |
+| --- | --- | --- |
+| `validity` | `-1.0` if syntax error, `+1.0` if tests fail, `0.0` if tests still pass | Rewards executable bugs, rejects broken syntax. |
+| `plausibility` | AST similarity score from `compute_ast_distance` | Rewards small realistic edits over random corruption. |
+| `learnability` | `+1.0` when recent solver success rate is between `0.1` and `0.9` | Rewards bugs that are neither trivial nor impossible. |
+
+The proposer is therefore rewarded for bugs that are:
+
+- valid Python,
+- test-breaking,
+- close to the original AST,
+- useful training examples for the solver.
+
+### Solver Reward
+
+Implemented in [`training/rewards.py`](training/rewards.py):
+
+```python
+solved = tests_passed and not syntax_error
+reward = 1.0 if solved else 0.0
+```
+
+Every solver result is recorded in a per-seed rolling deque of length `20`. The proposer uses this history through `get_solve_rate(seed_id)` to estimate whether a bug is learnable.
+
+### Plausibility Grader
+
+Implemented in [`server/plausibility.py`](server/plausibility.py).
+
+The plausibility score compares AST dumps of the clean and mutated code using a Levenshtein-style fuzz ratio:
+
+| AST similarity ratio | Score | Interpretation |
+| --- | --- | --- |
+| `100` | `0.0` | No edit, not a useful bug. |
+| `85` to `99` | `1.0` | Small realistic mutation. |
+| `50` to `84` | Linear decay down to `0.1` | Medium-sized change. |
+| `< 50` | `0.0` | Too different, likely unrealistic. |
+
+This discourages the proposer from replacing the whole function with nonsense.
+
+### Notebook Reward
+
+The Colab notebook at [`notebooks/train_colab.ipynb`](notebooks/train_colab.ipynb) uses the live OpenEnv server inside the reward function.
+
+For each model completion, it:
+
+1. Extracts Python code from the model output.
+2. Calls `env.reset()`.
+3. Calls `env.step(DebugzeroAction(...))`.
+4. Computes reward from the returned observation.
+
+That means training is connected to the real environment, not a static dataset. The notebook also evaluates baseline and trained policies and saves:
+
+- `results/reward_curve.png`
+- `results/loss_curve.png`
+- `results/baseline_vs_trained_reward.png`
+- `results/training_log.csv`
+
+## Bug Injection Logic
+
+The AST mutation engine lives in [`server/bug_injector.py`](server/bug_injector.py).
+
+`inject_bug(original_code, proposed_operator)` parses the clean code, applies one AST mutation, unparses the result, and accepts it only if all safety checks pass.
+
+Supported mutation operators:
+
+| Operator | Example behavior |
+| --- | --- |
+| `off_by_one` | Integer constants are shifted by `+1` or `-1`. |
+| `wrong_operator` | Comparisons and arithmetic operators are swapped, such as `<` to `>=` or `+` to `-`. |
+| `wrong_builtin` | Built-ins are swapped, such as `min`/`max`, `any`/`all`, or `sum`/`len`. |
+| `loop_boundary_shift` | `range(n)` becomes `range(n + 1)`, or a two-argument range shifts the start. |
+| `condition_negation` | `if condition` becomes `if not condition`. |
+| `missing_base_case` | A return inside an `if` body is replaced with `pass`. |
+| `slice_boundary_corruption` | Slice lower or upper bounds are shifted. |
+| `variable_swap` | Tuple assignment targets are swapped. |
+
+Accepted mutations must satisfy four checks:
+
+1. Original code parses.
+2. Mutated code is actually different.
+3. Mutated code does not include blocked imports.
+4. Mutated code parses after mutation.
+
+## Sandbox and Safety
+
+Execution is handled by [`server/executor.py`](server/executor.py).
+
+The executor builds:
+
+```python
+full_code = submitted_code + "\n\n" + tests
+```
+
+Then it validates and executes the code in a temporary file with a timeout.
+
+Safety checks include:
+
+- blocked imports: `os`, `sys`, `subprocess`, `shutil`, `pathlib`
+- blocked built-ins: `__import__`, `eval`, `exec`, `open`
+- AST parsing before execution
+- AST walk to catch direct `Import`, `ImportFrom`, and blocked function calls
+- subprocess timeout, currently `5` seconds
+- temporary directory isolation for each execution
+
+If code is unsafe but parses, the executor returns:
+
+```text
+Unsafe import detected.
+```
+
+If code does not parse, the executor returns a syntax-error observation.
+
+## Training Pipeline
+
+There are two training paths.
+
+### Recommended: Colab Notebook
+
+Use [`notebooks/train_colab.ipynb`](notebooks/train_colab.ipynb) for the hackathon submission.
+
+It:
+
+1. Installs DebugZero from GitHub.
+2. Starts the packaged OpenEnv FastAPI server, or connects to a remote HF Space URL.
+3. Runs an OpenEnv smoke test through `DebugzeroEnv`.
+4. Builds prompts from live environment resets.
+5. Uses TRL `GRPOTrainer`.
+6. Uses Unsloth when available, with native TRL fallback.
+7. Computes rewards through live `reset` and `step` calls.
+8. Saves plots for the README and final presentation.
+
+### Experimental Script
+
+[`training/grpo_train.py`](training/grpo_train.py) contains an experimental GRPO trainer configuration and the richer reward functions from [`training/rewards.py`](training/rewards.py). It is useful as implementation reference, but the notebook is the clearer end-to-end artifact for judges because it connects directly to the environment and saves visible training evidence.
+
+## Prompt Templates
+
+[`training/dual_role_sampler.py`](training/dual_role_sampler.py) defines two role prompts.
+
+The proposer prompt asks the model to:
+
+- inject an adversarial but plausible bug,
+- keep code syntax-valid,
+- make the function fail tests,
+- return only modified code.
+
+The solver prompt asks the model to:
+
+- inspect buggy code,
+- repair it,
+- return only corrected code.
+
+## Evaluation
+
+Tests live under [`eval/`](eval/).
+
+Current checks cover:
+
+- AST mutation behavior:
+  - missing base case,
+  - off-by-one mutation,
+  - loop boundary shift,
+  - wrong built-in,
+  - condition negation,
+  - safety checks.
+- Executor behavior:
+  - safe code passes,
+  - blocked imports are rejected,
+  - syntax errors are rejected,
+  - correct code passes tests,
+  - buggy code fails tests.
+
+There is also a plausibility evaluation scaffold in [`eval/plausibility_eval.py`](eval/plausibility_eval.py) for comparing generated bugs with human-like bugs from the navidadkhah dataset.
+
+Run local checks with:
 
-First, install dependencies:
 ```bash
-uv sync
+pytest eval
 ```
 
-## Running the Training Loop
+## Running Locally
 
-We provide a direct implementation of the Hugging Face TRL GRPOTrainer, modified to support a dual-role `reward_fn` dynamic routing system.
+Install dependencies:
 
 ```bash
-python training/grpo_train.py
+uv sync
 ```
 
-For the hackathon submission, use the rerunnable Colab notebook at
-[`notebooks/train_colab.ipynb`](notebooks/train_colab.ipynb). It installs DebugZero
-from GitHub, connects through the packaged OpenEnv client, trains with TRL/Unsloth
-against live `reset`/`step` environment rollouts, and saves reward/loss plots for
-the README.
+Start the OpenEnv server:
 
-## Repository Structure
+```bash
+python -m uvicorn server.app:app --host 0.0.0.0 --port 8000
+```
+
+Smoke-test with the client:
 
+```python
+from debugZero.client import DebugzeroEnv
+from debugZero.models import DebugzeroAction
+
+with DebugzeroEnv(base_url="http://localhost:8000") as env:
+    obs = env.reset().observation
+    print(obs.role_next)
+    print(obs.current_code)
+
+    buggy = obs.current_code.replace("distance < threshold", "distance <= threshold")
+    result = env.step(DebugzeroAction(role="proposer", code=buggy))
+    print(result.observation.tests_passed)
 ```
+
+## Repository Structure
+
+```text
 debugZero/
-├���─ server/
-│   ├── app.py                   # OpenEnv FastAPI app
-│   ├── debugZero_environment.py # Environment state machine tracking Proposer/Solver turns
-│   ├── executor.py              # Isolated subprocess execution sandbox (timeout=5s)
-│   ├── bug_injector.py          # AST mutation engine with 8 operators and strict safety checks
-│   └── plausibility.py          # Levenshtein AST distance calculator 
-├── training/
-│   ├── rewards.py               # Calculation for Proposer and Solver rewards + learnability deque
-│   ├── dual_role_sampler.py     # Prompt templates for Qwen2.5-Coder-3B-Instruct
-│   └── grpo_train.py            # TRL trainer bridging output metadata to reward forms
-├── eval/
-│   └── plausibility_eval.py     # Offline benchmark tool against navidadkhah 25k bug dataset
-├── models.py                    # Data types for the dual-role protocol
-└── client.py                    # OpenEnv DebugZero client class
+|-- openenv.yaml                  # OpenEnv manifest
+|-- README.md                     # Project and submission documentation
+|-- models.py                     # Action, observation, and state schemas
+|-- client.py                     # OpenEnv client
+|-- server/
+|   |-- app.py                    # FastAPI OpenEnv app
+|   |-- debugZero_environment.py  # Environment state machine
+|   |-- executor.py               # Code execution and safety checks
+|   |-- bug_injector.py           # AST mutation engine
+|   |-- plausibility.py           # AST similarity grader
+|   `-- requirements.txt          # HF Space server dependencies
+|-- training/
+|   |-- rewards.py                # Proposer and solver reward functions
+|   |-- dual_role_sampler.py      # Prompt templates
+|   `-- grpo_train.py             # Experimental GRPO trainer script
+|-- notebooks/
+|   `-- train_colab.ipynb         # Recommended rerunnable training notebook
+`-- eval/
+    |-- test_bug_injector.py      # Mutation tests
+    |-- test_executor.py          # Executor tests
+    `-- plausibility_eval.py      # Plausibility evaluation scaffold
+```
+
+## Deployment Notes
+
+The HF Space runs `server.app:app`, so imports are written to support both:
+
+- top-level Space import mode: `server.app`
+- installed package mode: `debugZero.server.app`
+
+Server dependencies for the Space are in [`server/requirements.txt`](server/requirements.txt). The server requires `thefuzz` because `server/plausibility.py` imports it during app startup.
+
+Because the Docker Space serves Uvicorn on port `8000`, the Hugging Face README metadata must include:
+
+```yaml
+sdk: docker
+app_port: 8000
 ```
 
-## Safety
+After pushing to Hugging Face, confirm:
+
+- the Space builds successfully,
+- `/schema` returns a valid OpenEnv schema,
+- `reset` returns the HumanEval seed code,
+- `step` returns `tests_passed` and `syntax_error`,
+- the README links to the final Space URL and training evidence.
+
+## Current Limitations and Next Steps
+
+Current implementation details to be aware of:
+
+- The server seed is currently a single HumanEval-style function, `HumanEval/0`.
+- The server verifies behavior but does not emit shaped scalar rewards yet. Training computes those externally from observations.
+- Tests are currently bundled in the environment seed. For a stronger benchmark, split public and hidden tests.
+- The AST bug injector exists as a utility, while proposer actions currently submit full mutated code.
+- The training notebook is the preferred proof artifact because it uses the live OpenEnv path and produces plots.
+
+High-impact next steps:
 
-DebugZero wraps the execution in a standard `openenv` container via FastAPI, but specifically filters AST payloads using an `is_safe` string filter to prevent `os`, `sys`, `subprocess`, `shutil`, and `pathlib` evaluations.
+- Add more HumanEval or curated seed tasks.
+- Move shaped reward metadata into observations for easier external analysis.
+- Add hidden tests and baseline-vs-trained examples to the README.
+- Use the AST injector to generate proposer warm-start examples.
+- Record qualitative before/after solver repairs for the final presentation.

graphify-out/GRAPH_REPORT.md

@@ -1,11 +1,11 @@
 # Graph Report - C:\Users\astra\Desktop\hackon\debugZero  (2026-04-25)
 
 ## Corpus Check
-- 16 files · ~11,370 words
+- 16 files · ~13,482 words
 - Verdict: corpus is large enough that graph structure adds value.
 
 ## Summary
-- 85 nodes · 139 edges · 11 communities detected
+- 85 nodes · 139 edges · 12 communities detected
 - Extraction: 60% EXTRACTED · 40% INFERRED · 0% AMBIGUOUS · INFERRED: 55 edges (avg confidence: 0.62)
 - Token cost: 0 input · 0 output
 
@@ -21,6 +21,7 @@
 - [[_COMMUNITY_Community 8|Community 8]]
 - [[_COMMUNITY_Community 9|Community 9]]
 - [[_COMMUNITY_Community 10|Community 10]]
+- [[_COMMUNITY_Community 11|Community 11]]
 
 ## God Nodes (most connected - your core abstractions)
 1. `DebugzeroObservation` - 16 edges
@@ -35,15 +36,15 @@
 10. `test_local_env()` - 5 edges
 
 ## Surprising Connections (you probably didn't know these)
-- `DebugzeroEnv` --uses--> `DebugzeroAction`  [INFERRED]
+- `DebugzeroEnv` --uses--> `DebugzeroObservation`  [INFERRED]
   C:\Users\astra\Desktop\hackon\debugZero\client.py → C:\Users\astra\Desktop\hackon\debugZero\models.py
 - `DebugzeroEnv` --uses--> `DebugzeroState`  [INFERRED]
   C:\Users\astra\Desktop\hackon\debugZero\client.py → C:\Users\astra\Desktop\hackon\debugZero\models.py
-- `Client for the DebugZero Environment.      This client maintains a persistent We` --uses--> `DebugzeroAction`  [INFERRED]
+- `Client for the DebugZero Environment.      This client maintains a persistent We` --uses--> `DebugzeroObservation`  [INFERRED]
   C:\Users\astra\Desktop\hackon\debugZero\client.py → C:\Users\astra\Desktop\hackon\debugZero\models.py
 - `Client for the DebugZero Environment.      This client maintains a persistent We` --uses--> `DebugzeroState`  [INFERRED]
   C:\Users\astra\Desktop\hackon\debugZero\client.py → C:\Users\astra\Desktop\hackon\debugZero\models.py
-- `Convert DebugzeroAction to JSON payload for step message.          Args:` --uses--> `DebugzeroAction`  [INFERRED]
+- `Convert DebugzeroAction to JSON payload for step message.          Args:` --uses--> `DebugzeroObservation`  [INFERRED]
   C:\Users\astra\Desktop\hackon\debugZero\client.py → C:\Users\astra\Desktop\hackon\debugZero\models.py
 
 ## Communities
@@ -57,38 +58,42 @@ Cohesion: 0.2
 Nodes (1): BugInjectorVisitor
 
 ### Community 2 - "Community 2"
-Cohesion: 0.33
-Nodes (7): DebugzeroEnv, Client for the DebugZero Environment.      This client maintains a persistent We, Convert DebugzeroAction to JSON payload for step message.          Args:, Parse server response into StepResult[DebugzeroObservation].          Args:, DebugzeroObservation, Observation from the DebugZero environment following sandbox execution., Observation
-
-### Community 3 - "Community 3"
 Cohesion: 0.28
 Nodes (2): main(), Entry point for direct execution via uv run or python -m.      This function ena
 
-### Community 4 - "Community 4"
+### Community 3 - "Community 3"
 Cohesion: 0.33
 Nodes (7): create_dataset(), main(), reward_fn(), compute_proposer_reward(), compute_solver_reward(), get_solve_rate(), record_solve_result()
 
-### Community 5 - "Community 5"
-Cohesion: 0.29
-Nodes (7): Action, Entry point for direct execution via uv run or python -m.      This function ena, Dual-role DebugZero Environment wrapping a Python sandbox execution     for Prop, Dual-role DebugZero Environment wrapping a Python sandbox execution     for Prop, ExecutionResult, DebugzeroAction, Action for the DebugZero environment representing the Proposer or Solver inputs.
+### Community 4 - "Community 4"
+Cohesion: 0.32
+Nodes (7): Action, Entry point for direct execution via uv run or python -m.      This function ena, DebugzeroEnv, Client for the DebugZero Environment.      This client maintains a persistent We, Convert DebugzeroAction to JSON payload for step message.          Args:, DebugzeroAction, Action for the DebugZero environment representing the Proposer or Solver inputs.
 
-### Community 6 - "Community 6"
+### Community 5 - "Community 5"
 Cohesion: 0.32
 Nodes (6): execute_code(), is_safe(), Check if the code contains any blocked imports strings.      Also performs a qu, Executes the provided python code alongside its tests in an isolated subprocess., test_execute_code(), test_executor_is_safe()
 
+### Community 6 - "Community 6"
+Cohesion: 0.4
+Nodes (4): Parse server response into State object.          Args:             payload: JSO, DebugzeroState, State for the DebugZero environment, extending default state with seed context., State
+
 ### Community 7 - "Community 7"
 Cohesion: 0.47
 Nodes (3): DebugzeroEnvironment, Environment, test_local_env()
 
 ### Community 8 - "Community 8"
-Cohesion: 0.4
-Nodes (4): Parse server response into State object.          Args:             payload: JSO, DebugzeroState, State for the DebugZero environment, extending default state with seed context., State
-
-### Community 9 - "Community 9"
 Cohesion: 0.33
 Nodes (4): compute_ast_distance(), evaluate_navidadkhah_plausibility(), Offline evaluation of generated bugs against the navidadkhah 25k bug dataset., Computes the string similarity distance between the AST dumps of the original
 
+### Community 9 - "Community 9"
+Cohesion: 0.5
+Nodes (4): Parse server response into StepResult[DebugzeroObservation].          Args:, DebugzeroObservation, Observation from the DebugZero environment following sandbox execution., Observation
+
 ### Community 10 - "Community 10"
+Cohesion: 0.5
+Nodes (3): Dual-role DebugZero Environment wrapping a Python sandbox execution     for Prop, Dual-role DebugZero Environment wrapping a Python sandbox execution     for Prop, ExecutionResult
+
+### Community 11 - "Community 11"
 Cohesion: 0.67
 Nodes (0): 
 
@@ -99,7 +104,7 @@ Nodes (0):
 ## Suggested Questions
 _Questions this graph is uniquely positioned to answer:_
 
-- **Why does `DebugzeroEnvironment` connect `Community 7` to `Community 8`, `Community 2`, `Community 3`, `Community 5`?**
+- **Why does `DebugzeroEnvironment` connect `Community 7` to `Community 2`, `Community 4`, `Community 6`, `Community 9`, `Community 10`?**
   _High betweenness centrality (0.213) - this node is a cross-community bridge._
 - **Why does `BugInjectorVisitor` connect `Community 1` to `Community 0`?**
   _High betweenness centrality (0.173) - this node is a cross-community bridge._

graphify-out/cache/28c79dacba9b7f6e353406b2afa843edb89da380af82cb43906368decabb8bb9.json

@@ -0,0 +1 @@
+{"nodes": [{"id": "c_users_astra_desktop_hackon_debugzero_server_app_py", "label": "app.py", "file_type": "code", "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\app.py", "source_location": "L1"}, {"id": "app_main", "label": "main()", "file_type": "code", "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\app.py", "source_location": "L63"}, {"id": "app_rationale_64", "label": "Entry point for direct execution via uv run or python -m.      This function ena", "file_type": "rationale", "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\app.py", "source_location": "L64"}], "edges": [{"source": "c_users_astra_desktop_hackon_debugzero_server_app_py", "target": "os", "relation": "imports", "confidence": "EXTRACTED", "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\app.py", "source_location": "L30", "weight": 1.0}, {"source": "c_users_astra_desktop_hackon_debugzero_server_app_py", "target": "openenv_core_env_server_http_server", "relation": "imports_from", "confidence": "EXTRACTED", "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\app.py", "source_location": "L36", "weight": 1.0}, {"source": "c_users_astra_desktop_hackon_debugzero_server_app_py", "target": "models", "relation": "imports_from", "confidence": "EXTRACTED", "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\app.py", "source_location": "L43", "weight": 1.0}, {"source": "c_users_astra_desktop_hackon_debugzero_server_app_py", "target": "c_users_astra_desktop_hackon_debugzero_server_debugzero_environment_py", "relation": "imports_from", "confidence": "EXTRACTED", "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\app.py", "source_location": "L44", "weight": 1.0}, {"source": "c_users_astra_desktop_hackon_debugzero_server_app_py", "target": "c_users_astra_desktop_hackon_debugzero_models_py", "relation": "imports_from", "confidence": "EXTRACTED", "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\app.py", "source_location": "L46", "weight": 1.0}, {"source": "c_users_astra_desktop_hackon_debugzero_server_app_py", "target": "c_users_astra_desktop_hackon_debugzero_server_debugzero_environment_py", "relation": "imports_from", "confidence": "EXTRACTED", "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\app.py", "source_location": "L47", "weight": 1.0}, {"source": "c_users_astra_desktop_hackon_debugzero_server_app_py", "target": "models", "relation": "imports_from", "confidence": "EXTRACTED", "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\app.py", "source_location": "L49", "weight": 1.0}, {"source": "c_users_astra_desktop_hackon_debugzero_server_app_py", "target": "server_debugzero_environment", "relation": "imports_from", "confidence": "EXTRACTED", "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\app.py", "source_location": "L50", "weight": 1.0}, {"source": "c_users_astra_desktop_hackon_debugzero_server_app_py", "target": "app_main", "relation": "contains", "confidence": "EXTRACTED", "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\app.py", "source_location": "L63", "weight": 1.0}, {"source": "app_rationale_64", "target": "app_main", "relation": "rationale_for", "confidence": "EXTRACTED", "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\app.py", "source_location": "L64", "weight": 1.0}], "raw_calls": [{"caller_nid": "app_main", "callee": "ArgumentParser", "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\app.py", "source_location": "L75"}, {"caller_nid": "app_main", "callee": "add_argument", "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\app.py", "source_location": "L76"}, {"caller_nid": "app_main", "callee": "add_argument", "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\app.py", "source_location": "L77"}, {"caller_nid": "app_main", "callee": "parse_args", "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\app.py", "source_location": "L78"}, {"caller_nid": "app_main", "callee": "run", "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\app.py", "source_location": "L80"}]}

graphify-out/graph.json

@@ -9,7 +9,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\client.py",
       "source_location": "L1",
       "id": "c_users_astra_desktop_hackon_debugzero_client_py",
-      "community": 3,
+      "community": 2,
       "norm_label": "client.py"
     },
     {
@@ -18,7 +18,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\client.py",
       "source_location": "L18",
       "id": "client_debugzeroenv",
-      "community": 2,
+      "community": 4,
       "norm_label": "debugzeroenv"
     },
     {
@@ -27,7 +27,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\client.py",
       "source_location": "L47",
       "id": "client_debugzeroenv_step_payload",
-      "community": 2,
+      "community": 4,
       "norm_label": "._step_payload()"
     },
     {
@@ -36,7 +36,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\client.py",
       "source_location": "L62",
       "id": "client_debugzeroenv_parse_result",
-      "community": 2,
+      "community": 9,
       "norm_label": "._parse_result()"
     },
     {
@@ -45,7 +45,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\client.py",
       "source_location": "L90",
       "id": "client_debugzeroenv_parse_state",
-      "community": 8,
+      "community": 6,
       "norm_label": "._parse_state()"
     },
     {
@@ -53,7 +53,7 @@
       "file_type": "rationale",
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\client.py",
       "source_location": "L21",
-      "community": 2,
+      "community": 4,
       "norm_label": "client for the debugzero environment.      this client maintains a persistent we",
       "id": "client_rationale_21"
     },
@@ -62,7 +62,7 @@
       "file_type": "rationale",
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\client.py",
       "source_location": "L48",
-      "community": 2,
+      "community": 4,
       "norm_label": "convert debugzeroaction to json payload for step message.          args:",
       "id": "client_rationale_48"
     },
@@ -71,7 +71,7 @@
       "file_type": "rationale",
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\client.py",
       "source_location": "L63",
-      "community": 2,
+      "community": 9,
       "norm_label": "parse server response into stepresult[debugzeroobservation].          args:",
       "id": "client_rationale_63"
     },
@@ -80,7 +80,7 @@
       "file_type": "rationale",
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\client.py",
       "source_location": "L91",
-      "community": 8,
+      "community": 6,
       "norm_label": "parse server response into state object.          args:             payload: jso",
       "id": "client_rationale_91"
     },
@@ -90,7 +90,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\models.py",
       "source_location": "L1",
       "id": "c_users_astra_desktop_hackon_debugzero_models_py",
-      "community": 3,
+      "community": 2,
       "norm_label": "models.py"
     },
     {
@@ -99,7 +99,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\models.py",
       "source_location": "L18",
       "id": "models_debugzeroaction",
-      "community": 5,
+      "community": 4,
       "norm_label": "debugzeroaction"
     },
     {
@@ -108,7 +108,7 @@
       "source_file": "",
       "source_location": "",
       "id": "action",
-      "community": 5,
+      "community": 4,
       "norm_label": "action"
     },
     {
@@ -117,7 +117,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\models.py",
       "source_location": "L25",
       "id": "models_debugzeroobservation",
-      "community": 2,
+      "community": 9,
       "norm_label": "debugzeroobservation"
     },
     {
@@ -126,7 +126,7 @@
       "source_file": "",
       "source_location": "",
       "id": "observation",
-      "community": 2,
+      "community": 9,
       "norm_label": "observation"
     },
     {
@@ -135,7 +135,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\models.py",
       "source_location": "L34",
       "id": "models_debugzerostate",
-      "community": 8,
+      "community": 6,
       "norm_label": "debugzerostate"
     },
     {
@@ -144,7 +144,7 @@
       "source_file": "",
       "source_location": "",
       "id": "state",
-      "community": 8,
+      "community": 6,
       "norm_label": "state"
     },
     {
@@ -152,7 +152,7 @@
       "file_type": "rationale",
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\models.py",
       "source_location": "L19",
-      "community": 5,
+      "community": 4,
       "norm_label": "action for the debugzero environment representing the proposer or solver inputs.",
       "id": "models_rationale_19"
     },
@@ -161,7 +161,7 @@
       "file_type": "rationale",
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\models.py",
       "source_location": "L26",
-      "community": 2,
+      "community": 9,
       "norm_label": "observation from the debugzero environment following sandbox execution.",
       "id": "models_rationale_26"
     },
@@ -170,7 +170,7 @@
       "file_type": "rationale",
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\models.py",
       "source_location": "L35",
-      "community": 8,
+      "community": 6,
       "norm_label": "state for the debugzero environment, extending default state with seed context.",
       "id": "models_rationale_35"
     },
@@ -198,7 +198,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\__init__.py",
       "source_location": "L1",
       "id": "c_users_astra_desktop_hackon_debugzero_init_py",
-      "community": 3,
+      "community": 2,
       "norm_label": "__init__.py"
     },
     {
@@ -207,7 +207,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\eval\\plausibility_eval.py",
       "source_location": "L1",
       "id": "c_users_astra_desktop_hackon_debugzero_eval_plausibility_eval_py",
-      "community": 9,
+      "community": 8,
       "norm_label": "plausibility_eval.py"
     },
     {
@@ -216,7 +216,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\eval\\plausibility_eval.py",
       "source_location": "L5",
       "id": "plausibility_eval_evaluate_navidadkhah_plausibility",
-      "community": 9,
+      "community": 8,
       "norm_label": "evaluate_navidadkhah_plausibility()"
     },
     {
@@ -224,7 +224,7 @@
       "file_type": "rationale",
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\eval\\plausibility_eval.py",
       "source_location": "L6",
-      "community": 9,
+      "community": 8,
       "norm_label": "offline evaluation of generated bugs against the navidadkhah 25k bug dataset.",
       "id": "plausibility_eval_rationale_6"
     },
@@ -297,7 +297,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\eval\\test_executor.py",
       "source_location": "L1",
       "id": "c_users_astra_desktop_hackon_debugzero_eval_test_executor_py",
-      "community": 6,
+      "community": 5,
       "norm_label": "test_executor.py"
     },
     {
@@ -306,7 +306,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\eval\\test_executor.py",
       "source_location": "L4",
       "id": "test_executor_test_executor_is_safe",
-      "community": 6,
+      "community": 5,
       "norm_label": "test_executor_is_safe()"
     },
     {
@@ -315,7 +315,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\eval\\test_executor.py",
       "source_location": "L16",
       "id": "test_executor_test_execute_code",
-      "community": 6,
+      "community": 5,
       "norm_label": "test_execute_code()"
     },
     {
@@ -324,7 +324,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\app.py",
       "source_location": "L1",
       "id": "c_users_astra_desktop_hackon_debugzero_server_app_py",
-      "community": 3,
+      "community": 2,
       "norm_label": "app.py"
     },
     {
@@ -333,7 +333,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\app.py",
       "source_location": "L63",
       "id": "app_main",
-      "community": 3,
+      "community": 2,
       "norm_label": "main()"
     },
     {
@@ -341,9 +341,9 @@
       "file_type": "rationale",
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\app.py",
       "source_location": "L64",
-      "id": "app_rationale_64",
-      "community": 3,
-      "norm_label": "entry point for direct execution via uv run or python -m.      this function ena"
+      "community": 2,
+      "norm_label": "entry point for direct execution via uv run or python -m.      this function ena",
+      "id": "app_rationale_64"
     },
     {
       "label": "bug_injector.py",
@@ -477,7 +477,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\debugZero_environment.py",
       "source_location": "L1",
       "id": "c_users_astra_desktop_hackon_debugzero_server_debugzero_environment_py",
-      "community": 3,
+      "community": 2,
       "norm_label": "debugzero_environment.py"
     },
     {
@@ -504,7 +504,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\debugZero_environment.py",
       "source_location": "L47",
       "id": "debugzero_environment_debugzeroenvironment_init",
-      "community": 8,
+      "community": 6,
       "norm_label": ".__init__()"
     },
     {
@@ -531,7 +531,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\debugZero_environment.py",
       "source_location": "L129",
       "id": "debugzero_environment_state",
-      "community": 3,
+      "community": 2,
       "norm_label": "state()"
     },
     {
@@ -539,9 +539,9 @@
       "file_type": "rationale",
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\debugZero_environment.py",
       "source_location": "L41",
-      "id": "debugzero_environment_rationale_41",
-      "community": 5,
-      "norm_label": "dual-role debugzero environment wrapping a python sandbox execution     for prop"
+      "community": 10,
+      "norm_label": "dual-role debugzero environment wrapping a python sandbox execution     for prop",
+      "id": "debugzero_environment_rationale_41"
     },
     {
       "label": "executor.py",
@@ -549,7 +549,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\executor.py",
       "source_location": "L1",
       "id": "c_users_astra_desktop_hackon_debugzero_server_executor_py",
-      "community": 6,
+      "community": 5,
       "norm_label": "executor.py"
     },
     {
@@ -558,7 +558,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\executor.py",
       "source_location": "L9",
       "id": "executor_is_safe",
-      "community": 6,
+      "community": 5,
       "norm_label": "is_safe()"
     },
     {
@@ -567,7 +567,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\executor.py",
       "source_location": "L44",
       "id": "executor_executionresult",
-      "community": 5,
+      "community": 10,
       "norm_label": "executionresult"
     },
     {
@@ -576,7 +576,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\executor.py",
       "source_location": "L45",
       "id": "executor_executionresult_init",
-      "community": 5,
+      "community": 10,
       "norm_label": ".__init__()"
     },
     {
@@ -585,7 +585,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\executor.py",
       "source_location": "L51",
       "id": "executor_execute_code",
-      "community": 6,
+      "community": 5,
       "norm_label": "execute_code()"
     },
     {
@@ -593,7 +593,7 @@
       "file_type": "rationale",
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\executor.py",
       "source_location": "L10",
-      "community": 6,
+      "community": 5,
       "norm_label": "check if the code contains any blocked imports strings.      also performs a qu",
       "id": "executor_rationale_10"
     },
@@ -602,7 +602,7 @@
       "file_type": "rationale",
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\executor.py",
       "source_location": "L52",
-      "community": 6,
+      "community": 5,
       "norm_label": "executes the provided python code alongside its tests in an isolated subprocess.",
       "id": "executor_rationale_52"
     },
@@ -612,7 +612,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\plausibility.py",
       "source_location": "L1",
       "id": "c_users_astra_desktop_hackon_debugzero_server_plausibility_py",
-      "community": 9,
+      "community": 8,
       "norm_label": "plausibility.py"
     },
     {
@@ -621,7 +621,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\plausibility.py",
       "source_location": "L4",
       "id": "plausibility_compute_ast_distance",
-      "community": 9,
+      "community": 8,
       "norm_label": "compute_ast_distance()"
     },
     {
@@ -629,7 +629,7 @@
       "file_type": "rationale",
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\plausibility.py",
       "source_location": "L5",
-      "community": 9,
+      "community": 8,
       "norm_label": "computes the string similarity distance between the ast dumps of the original",
       "id": "plausibility_rationale_5"
     },
@@ -639,7 +639,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\__init__.py",
       "source_location": "L1",
       "id": "c_users_astra_desktop_hackon_debugzero_server_init_py",
-      "community": 3,
+      "community": 2,
       "norm_label": "__init__.py"
     },
     {
@@ -648,7 +648,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\training\\dual_role_sampler.py",
       "source_location": "L1",
       "id": "c_users_astra_desktop_hackon_debugzero_training_dual_role_sampler_py",
-      "community": 10,
+      "community": 11,
       "norm_label": "dual_role_sampler.py"
     },
     {
@@ -657,7 +657,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\training\\dual_role_sampler.py",
       "source_location": "L23",
       "id": "dual_role_sampler_sample_proposer_prompt",
-      "community": 10,
+      "community": 11,
       "norm_label": "sample_proposer_prompt()"
     },
     {
@@ -666,7 +666,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\training\\dual_role_sampler.py",
       "source_location": "L26",
       "id": "dual_role_sampler_sample_solver_prompt",
-      "community": 10,
+      "community": 11,
       "norm_label": "sample_solver_prompt()"
     },
     {
@@ -675,7 +675,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\training\\grpo_train.py",
       "source_location": "L1",
       "id": "c_users_astra_desktop_hackon_debugzero_training_grpo_train_py",
-      "community": 4,
+      "community": 3,
       "norm_label": "grpo_train.py"
     },
     {
@@ -684,7 +684,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\training\\grpo_train.py",
       "source_location": "L25",
       "id": "grpo_train_reward_fn",
-      "community": 4,
+      "community": 3,
       "norm_label": "reward_fn()"
     },
     {
@@ -693,7 +693,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\training\\grpo_train.py",
       "source_location": "L44",
       "id": "grpo_train_create_dataset",
-      "community": 4,
+      "community": 3,
       "norm_label": "create_dataset()"
     },
     {
@@ -702,7 +702,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\training\\grpo_train.py",
       "source_location": "L57",
       "id": "grpo_train_main",
-      "community": 4,
+      "community": 3,
       "norm_label": "main()"
     },
     {
@@ -711,7 +711,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\training\\rewards.py",
       "source_location": "L1",
       "id": "c_users_astra_desktop_hackon_debugzero_training_rewards_py",
-      "community": 4,
+      "community": 3,
       "norm_label": "rewards.py"
     },
     {
@@ -720,7 +720,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\training\\rewards.py",
       "source_location": "L7",
       "id": "rewards_get_solve_rate",
-      "community": 4,
+      "community": 3,
       "norm_label": "get_solve_rate()"
     },
     {
@@ -729,7 +729,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\training\\rewards.py",
       "source_location": "L13",
       "id": "rewards_record_solve_result",
-      "community": 4,
+      "community": 3,
       "norm_label": "record_solve_result()"
     },
     {
@@ -738,7 +738,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\training\\rewards.py",
       "source_location": "L18",
       "id": "rewards_compute_proposer_reward",
-      "community": 4,
+      "community": 3,
       "norm_label": "compute_proposer_reward()"
     },
     {
@@ -747,7 +747,7 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\training\\rewards.py",
       "source_location": "L37",
       "id": "rewards_compute_solver_reward",
-      "community": 4,
+      "community": 3,
       "norm_label": "compute_solver_reward()"
     },
     {
@@ -755,7 +755,7 @@
       "file_type": "rationale",
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\app.py",
       "source_location": "L61",
-      "community": 5,
+      "community": 4,
       "norm_label": "entry point for direct execution via uv run or python -m.      this function ena",
       "id": "app_rationale_61"
     },
@@ -764,7 +764,7 @@
       "file_type": "rationale",
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\debugZero_environment.py",
       "source_location": "L39",
-      "community": 5,
+      "community": 10,
       "norm_label": "dual-role debugzero environment wrapping a python sandbox execution     for prop",
       "id": "debugzero_environment_rationale_39"
     }
@@ -1196,11 +1196,11 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\app.py",
       "source_location": "L49",
       "weight": 0.8,
-      "_src": "app_rationale_64",
-      "_tgt": "models_debugzeroaction",
+      "_src": "models_debugzeroaction",
+      "_tgt": "app_rationale_64",
+      "confidence_score": 0.5,
       "source": "models_debugzeroaction",
-      "target": "app_rationale_64",
-      "confidence_score": 0.5
+      "target": "app_rationale_64"
     },
     {
       "relation": "uses",
@@ -1220,11 +1220,11 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\debugZero_environment.py",
       "source_location": "L21",
       "weight": 0.8,
-      "_src": "debugzero_environment_rationale_41",
-      "_tgt": "models_debugzeroaction",
+      "_src": "models_debugzeroaction",
+      "_tgt": "debugzero_environment_rationale_41",
+      "confidence_score": 0.5,
       "source": "models_debugzeroaction",
-      "target": "debugzero_environment_rationale_41",
-      "confidence_score": 0.5
+      "target": "debugzero_environment_rationale_41"
     },
     {
       "relation": "calls",
@@ -1292,11 +1292,11 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\app.py",
       "source_location": "L49",
       "weight": 0.8,
-      "_src": "app_rationale_64",
-      "_tgt": "models_debugzeroobservation",
+      "_src": "models_debugzeroobservation",
+      "_tgt": "app_rationale_64",
+      "confidence_score": 0.5,
       "source": "models_debugzeroobservation",
-      "target": "app_rationale_64",
-      "confidence_score": 0.5
+      "target": "app_rationale_64"
     },
     {
       "relation": "uses",
@@ -1316,11 +1316,11 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\debugZero_environment.py",
       "source_location": "L21",
       "weight": 0.8,
-      "_src": "debugzero_environment_rationale_41",
-      "_tgt": "models_debugzeroobservation",
+      "_src": "models_debugzeroobservation",
+      "_tgt": "debugzero_environment_rationale_41",
+      "confidence_score": 0.5,
       "source": "models_debugzeroobservation",
-      "target": "debugzero_environment_rationale_41",
-      "confidence_score": 0.5
+      "target": "debugzero_environment_rationale_41"
     },
     {
       "relation": "calls",
@@ -1412,11 +1412,11 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\debugZero_environment.py",
       "source_location": "L21",
       "weight": 0.8,
-      "_src": "debugzero_environment_rationale_41",
-      "_tgt": "models_debugzerostate",
+      "_src": "models_debugzerostate",
+      "_tgt": "debugzero_environment_rationale_41",
+      "confidence_score": 0.5,
       "source": "models_debugzerostate",
-      "target": "debugzero_environment_rationale_41",
-      "confidence_score": 0.5
+      "target": "debugzero_environment_rationale_41"
     },
     {
       "relation": "calls",
@@ -1774,9 +1774,9 @@
       "weight": 0.8,
       "_src": "app_rationale_64",
       "_tgt": "debugzero_environment_debugzeroenvironment",
+      "confidence_score": 0.5,
       "source": "app_rationale_64",
-      "target": "debugzero_environment_debugzeroenvironment",
-      "confidence_score": 0.5
+      "target": "debugzero_environment_debugzeroenvironment"
     },
     {
       "relation": "contains",
@@ -2108,8 +2108,8 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\app.py",
       "source_location": "L47",
       "weight": 0.8,
-      "_src": "app_rationale_61",
-      "_tgt": "debugzero_environment_debugzeroenvironment",
+      "_src": "debugzero_environment_debugzeroenvironment",
+      "_tgt": "app_rationale_61",
       "confidence_score": 0.5,
       "source": "debugzero_environment_debugzeroenvironment",
       "target": "app_rationale_61"
@@ -2134,9 +2134,9 @@
       "weight": 0.8,
       "_src": "debugzero_environment_rationale_41",
       "_tgt": "executor_executionresult",
+      "confidence_score": 0.5,
       "source": "debugzero_environment_rationale_41",
-      "target": "executor_executionresult",
-      "confidence_score": 0.5
+      "target": "executor_executionresult"
     },
     {
       "relation": "contains",
@@ -2228,8 +2228,8 @@
       "source_file": "C:\\Users\\astra\\Desktop\\hackon\\debugZero\\server\\debugZero_environment.py",
       "source_location": "L26",
       "weight": 0.8,
-      "_src": "debugzero_environment_rationale_39",
-      "_tgt": "executor_executionresult",
+      "_src": "executor_executionresult",
+      "_tgt": "debugzero_environment_rationale_39",
       "confidence_score": 0.5,
       "source": "executor_executionresult",
       "target": "debugzero_environment_rationale_39"

openenv.yaml

@@ -5,4 +5,5 @@ type: space
 runtime: fastapi
 app: server.app:app
 port: 8000
-
+workers: 4
+max_concurrent_envs: 100

server/app.py

@@ -56,7 +56,7 @@ app = create_app(
     DebugzeroAction,
     DebugzeroObservation,
     env_name="debugZero",
-    max_concurrent_envs=1,  # increase this number to allow more concurrent WebSocket sessions
+    max_concurrent_envs=int(os.environ.get("MAX_CONCURRENT_ENVS", "100")),
 )