4.9.0 groundwork: correctness fixes, dead code removal, CI

.github/workflows/ci.yml

@@ -0,0 +1,52 @@
+name: CI
+
+on:
+  push:
+    branches: [main, refactor/4.9.0]
+  pull_request:
+
+jobs:
+  lint:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: astral-sh/setup-uv@v5
+      # Error-level rules only for now (syntax errors, undefined names,
+      # invalid prints/f-strings). Widen as the Phase 2 burndown lands.
+      - run: uvx ruff check --select E9,F63,F7,F82 gas neurons tests
+
+  test:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: astral-sh/setup-uv@v5
+      - name: Install locked dependencies (minus the GPU stack)
+        # Single source of truth: every version comes from uv.lock.
+        # The GPU stack (several GB of CUDA wheels) is skipped by NAME
+        # only — the current tests don't import torch. If a test ever
+        # needs it, drop the corresponding flag.
+        run: >
+          uv sync --frozen
+          --no-install-package torch
+          --no-install-package torchvision
+          --no-install-package torchaudio
+          --no-install-package torchcodec
+          --no-install-package triton
+          --no-install-package bitsandbytes
+          --no-install-package open-clip-torch
+          --no-install-package nvidia-cublas-cu12
+          --no-install-package nvidia-cuda-cupti-cu12
+          --no-install-package nvidia-cuda-nvrtc-cu12
+          --no-install-package nvidia-cuda-runtime-cu12
+          --no-install-package nvidia-cudnn-cu12
+          --no-install-package nvidia-cufft-cu12
+          --no-install-package nvidia-cufile-cu12
+          --no-install-package nvidia-curand-cu12
+          --no-install-package nvidia-cusolver-cu12
+          --no-install-package nvidia-cusparse-cu12
+          --no-install-package nvidia-cusparselt-cu12
+          --no-install-package nvidia-nccl-cu12
+          --no-install-package nvidia-nvjitlink-cu12
+          --no-install-package nvidia-nvtx-cu12
+      - name: Run tests
+        run: uv run --no-sync pytest tests/ -q

.gitignore

@@ -3,12 +3,6 @@ __pycache__/
 *.py[cod]
 *$py.class
 *.so
-
-# Node.js (not used - Python only project)
-node_modules/
-package.json
-package-lock.json
-.Python
 build/
 develop-eggs/
 dist/
@@ -27,17 +21,26 @@ wheels/
 .coverage
 *.ipynb
 
+# Node.js
+node_modules/
+package.json
+package-lock.json
+
 # Virtual environments
 venv/
 env/
 ENV/
 .venv/
+.Python
 
-# IDE
+# IDE / editors
 .vscode/
 .idea/
 *.swp
 *.swo
+\#*\#
+.#*
+*~
 
 # OS
 .DS_Store
@@ -46,39 +49,26 @@ Thumbs.db
 # Logs
 *.log
 
-# Environment variables
+# Environment variables (catch-all; templates stay tracked)
 .env
-.env.local
-.env.development
-.env.test
-.env.production
+.env.*
+*.env
+!*.env.template
+!.env.*.template
 
 # Model files and cache
 *.pth
 *.pt
 *.ckpt
+*.onnx
 .cache/
+downloaded_models/*
 
 # Bittensor specific
 ~/.bittensor/
-*.wallet 
-
-.env.validator
-.env.miner
-.env.generator
-.env.mock_miner
-
-*.onnx
-downloaded_models/*
+*.wallet
 
-.env.gen_miner
 base_miner/test_output/*
-base_miner/.env.gen_miner
-*.onnx
-downloaded_models/*
-
 generated_content/*
-
 wandb/*
-
-sample-images/ 
+sample-images/

Dockerfile

@@ -71,32 +71,9 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
     python3.10-dev \
     build-essential \
     ffmpeg \
-    xvfb \
     wget \
-    gnupg2 \
-    libnss3 \
-    libnspr4 \
-    libatk1.0-0 \
-    libatk-bridge2.0-0 \
-    libcups2 \
-    libxcomposite1 \
-    libxdamage1 \
-    libxrandr2 \
-    libgbm1 \
-    libpango-1.0-0 \
-    libcairo2 \
-    libasound2 \
-    libxshmfence1 \
     && rm -rf /var/lib/apt/lists/*
 
-# Install Google Chrome (needed by data service Selenium scraper)
-RUN wget -q -O - https://dl.google.com/linux/linux_signing_key.pub | apt-key add - && \
-    echo "deb [arch=amd64] http://dl.google.com/linux/chrome/deb/ stable main" \
-        > /etc/apt/sources.list.d/google-chrome.list && \
-    apt-get update && \
-    apt-get install -y --no-install-recommends google-chrome-stable && \
-    rm -rf /var/lib/apt/lists/*
-
 # Make python3.10 the default
 RUN update-alternatives --install /usr/bin/python3 python3 /usr/bin/python3.10 1 && \
     update-alternatives --install /usr/bin/python python /usr/bin/python3.10 1

docs/Installation.md

@@ -12,10 +12,9 @@ The installation script will automatically install the following system dependen
 
 - **Build tools**: pkg-config, cmake
 - **Media processing**: ffmpeg
-- **Browser automation**: Google Chrome, libnss3, libnspr4, xvfb
 - **Process management**: Node.js, npm, PM2, dotenv
 
-**Note**: The `--no-system-deps` option is primarily intended for **discriminative miners** who only need to submit models and don't require Chrome browser automation or Node.js process management tools. Validators should use the full installation to ensure all dependencies are available.
+**Note**: The `--no-system-deps` option is primarily intended for **discriminative miners** who only need to submit models and don't require Node.js process management tools. Validators should use the full installation to ensure all dependencies are available.
 
 ### Installing uv
 
@@ -49,7 +48,6 @@ The installation script will:
 - Check for Python 3.10+ and uv
 - Install system dependencies (unless `--no-system-deps` is specified):
   - pkg-config, cmake, ffmpeg
-  - chrome web driver, libnss3, libnspr4, xvfb
   - Node.js, npm, PM2, dotenv
 - Create a virtual environment using uv (fast dependency resolution)
 - Install all dependencies from `pyproject.toml`

gas/cache/db/connection.py

@@ -19,6 +19,7 @@ def __init__(self, db_path: Path):
     @contextlib.contextmanager
     def connect(self, max_retries: int = 3, retry_delay: float = 1.0) -> Iterator[sqlite3.Connection]:
         """Context manager for database connections with retry logic."""
+        conn = None
         for attempt in range(max_retries):
             try:
                 conn = sqlite3.connect(
@@ -30,11 +31,11 @@ def connect(self, max_retries: int = 3, retry_delay: float = 1.0) -> Iterator[sq
                 conn.execute("PRAGMA journal_mode=WAL")
                 conn.execute("PRAGMA busy_timeout=30000")
                 conn.execute("PRAGMA foreign_keys=ON")
-
-                yield conn
-                conn.close()
-                return
+                break
             except sqlite3.OperationalError as e:
+                if conn is not None:
+                    conn.close()
+                    conn = None
                 if "unable to open database file" in str(e) or "database is locked" in str(e):
                     if attempt < max_retries - 1:
                         bt.logging.warning(
@@ -46,9 +47,11 @@ def connect(self, max_retries: int = 3, retry_delay: float = 1.0) -> Iterator[sq
                     bt.logging.error(f"Failed to open database after {max_retries} attempts: {e}")
                     raise
                 raise
-            except Exception as e:
-                bt.logging.error(f"Database error: {e}")
-                raise
+
+        try:
+            yield conn
+        finally:
+            conn.close()
 
 
 def create_schema(conn: sqlite3.Connection) -> None:

gas/cache/db/media_store.py

@@ -360,23 +360,27 @@ def get_unuploaded_media(
                 conn.row_factory = __import__("sqlite3").Row
                 base_where = "(uploaded = 0 OR uploaded IS NULL)"
 
+                params: list = []
                 if source_type == 'miner':
                     source_filter = "(source_type = 'miner' AND verified = 1)"
                 elif source_type == 'generated':
                     source_filter = "source_type = 'generated'"
                 elif source_type in ('scraper', 'dataset'):
-                    source_filter = f"source_type = '{source_type}'"
+                    source_filter = "source_type = ?"
+                    params.append(source_type)
                 else:
                     source_filter = "(source_type IN ('scraper', 'dataset', 'generated') OR (source_type = 'miner' AND verified = 1))"
 
                 query = f"SELECT * FROM media WHERE {base_where} AND {source_filter}"
                 if modality:
-                    query += f" AND modality = '{modality}'"
+                    query += " AND modality = ?"
+                    params.append(modality)
                 query += " ORDER BY (source_type = 'miner' AND verified = 1) DESC, created_at ASC"
                 if limit is not None:
-                    query += f" LIMIT {limit}"
+                    query += " LIMIT ?"
+                    params.append(limit)
 
-                cursor = conn.execute(query)
+                cursor = conn.execute(query, params)
                 entries = []
                 for row in cursor.fetchall():
                     entry = self._row_to_media_entry(row)
@@ -717,7 +721,6 @@ def cleanup_uploaded_media(
                 prompt_ids = [r["prompt_id"] for r in rows if r["prompt_id"]]
 
                 mid_ph = ",".join("?" * len(media_ids))
-                conn.execute(f"DELETE FROM prompts WHERE id IN ({mid_ph})", media_ids)
                 conn.execute(f"DELETE FROM generator_challenge_outcomes WHERE media_id IN ({mid_ph})", media_ids)
                 conn.execute(f"DELETE FROM media WHERE id IN ({mid_ph})", media_ids)
                 conn.commit()

gas/cache/media_storage.py

@@ -335,7 +335,7 @@ def retrieve_media(
                     mask = None
                     mask_path = None
                     if kwargs.get("require_mask", False):
-                        mask_path = file_path.with_suffix("_mask.npy")
+                        mask_path = file_path.parent / (file_path.stem + "_mask.npy")
                         mask = self.read_mask(mask_path)
                         if mask is None:
                             print(f"Mask not found for {file_path}")

gas/generation/__init__.py

@@ -8,10 +8,28 @@
 
 Re-exports below preserve the historical flat import paths so external
 callers can keep using `from gas.generation import PromptGenerator` etc.
+They are resolved lazily (PEP 562) so that importing a pure-python
+submodule (e.g. gas.generation.prompts.prompt_qc) does not drag in the
+torch/diffusers/janus runtime stack.
 """
 
-from .media.generation_pipeline import GenerationPipeline
-from .media.models import initialize_model_registry
-from .prompts.model_prompt_styles import AdaptedPrompt, adapt_for_local_model
-from .prompts.prompt_generator import PromptGenerator
-from .prompts.scene import SceneDescription, extract_scene_with_vlm
+_LAZY_EXPORTS = {
+    "GenerationPipeline": "gas.generation.media.generation_pipeline",
+    "initialize_model_registry": "gas.generation.media.models",
+    "AdaptedPrompt": "gas.generation.prompts.model_prompt_styles",
+    "adapt_for_local_model": "gas.generation.prompts.model_prompt_styles",
+    "PromptGenerator": "gas.generation.prompts.prompt_generator",
+    "SceneDescription": "gas.generation.prompts.scene",
+    "extract_scene_with_vlm": "gas.generation.prompts.scene",
+}
+
+__all__ = list(_LAZY_EXPORTS)
+
+
+def __getattr__(name):
+    module_path = _LAZY_EXPORTS.get(name)
+    if module_path is None:
+        raise AttributeError(f"module {__name__!r} has no attribute {name!r}")
+    import importlib
+
+    return getattr(importlib.import_module(module_path), name)

gas/generation/media/generation_pipeline.py

@@ -240,12 +240,11 @@ def _generate_media_with_model(self, model_name, prompt, image):
 
         if task == "i2i" and image is None:
             raise ValueError(
-                "An image must be provided for image-to-image model {model_name}"
+                f"An image must be provided for image-to-image model {model_name}"
             )
 
         if not self._load_model(model_name):
             raise RuntimeError(f"Failed to load {model_name}")
-            return {}
 
         bt.logging.debug("Preparing generation arguments")
         gen_args = model_config.get("generation_args", {}).copy()