curie-backend 0.1.0__tar.gz

curie_backend-0.1.0/PKG-INFO

@@ -0,0 +1,116 @@
+Metadata-Version: 2.4
+Name: curie-backend
+Version: 0.1.0
+Summary: Text2SQL Assistant Backend
+Classifier: Programming Language :: Python :: 3
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+Requires-Dist: fastapi[standard]>=0.110.0
+Requires-Dist: uvicorn>=0.28.0
+Requires-Dist: torch>=2.1.0
+Requires-Dist: transformers>=4.38.0
+Requires-Dist: peft>=0.9.0
+Requires-Dist: accelerate>=0.27.0
+Requires-Dist: SQLAlchemy>=2.0.0
+Requires-Dist: psycopg[binary]>=3.1.0
+Requires-Dist: huggingface-hub>=0.21.0
+Requires-Dist: python-dotenv>=1.0.0
+
+# Project Curie - Text2SQL
+
+
+
+## Project description
+* **Idea:** Use natural language prompts to generate SQL queries 
+* **Front-end: VS Code Plugin**
+  * Integration of trained LLM into useful plugin with quick insights 
+  * Establish connection with existing database 
+  * Use natural language to describe queries to this database, which are generated by fine-tuned LLM 
+  * Ability to run generated SQL code directly via plugin to see results immediately with no bothersome copying
+
+* ...
+
+### Simple example: 
+![simple_example.png](other/diagrams/simple_example.png)
+
+## Project structure
+
+```text
+evaluation/   Benchmark setups for Text-to-SQL evaluation
+data/         Local datasets and sample data
+docs/         Project documentation, presentations, and research notes
+experiments/  Ongoing fine-tuning experiments and run notes
+extension/    VS Code extension and demo database
+ml/           Training, fine-tuning, and inference code
+ml/models/    Model metadata, configs, and model cards
+```
+
+## Project team and responsibilities
+
+| Name              | E-Mail                   | Identifier | Responsibilities / tasks |
+|-------------------|--------------------------|------------|--------------------------|
+| Tamer Lewis       | tamer.lewis@tum.de       | go68maj    | TBD                      |
+| Antonia Gilles    | antonia.gilles@tum.de    | go34cih    | TBD                      |
+| Leon Pfänder      | leon.pfaender@tum.de     | ge27fob    | TBD                      |
+| Christoph Klein   | christoph.m.klein@tum.de | ge42low    | TBD                      |
+
+
+
+
+## Project setup
+
+The project uses Python 3.12 and a pinned ML stack from `requirements.txt`.
+
+```bash
+python3.12 -m venv .venv
+source .venv/bin/activate
+python -m pip install --upgrade pip
+python -m pip install -r requirements.txt
+```
+
+Verify that the installed dependencies are consistent:
+
+```bash
+python -m pip check
+```
+
+### Git LFS for model weights
+
+Git LFS stores large model binaries on the GitLab LFS server while Git tracks
+small pointer files. This repository uses LFS for `*.safetensors` files.
+
+Minimal setup:
+
+```bash
+brew install git-lfs
+git lfs install
+git lfs pull
+```
+
+Add and push model weights through the normal Git workflow:
+
+```bash
+git add .gitattributes ml/models/<model>/adapter_model.safetensors
+git commit -m "Add model adapter"
+git push
+```
+
+After cloning, Git LFS normally downloads tracked files automatically. If a model file is only an LFS pointer or is missing, run `git lfs pull`. Verify the available model files with `git lfs ls-files`.
+
+LFS does not change how the benchmark is invoked:
+
+```bash
+python3 evaluation/spider/run.py \
+  --split dev \
+  --model-dir ml/models/<model>
+```
+
+### Spider
+
+Download the Spider dataset and place it in the `data/` folder:
+
+https://drive.google.com/file/d/1403EGqzIDoHMdQF4c9Bkyl7dZLZ5Wt6J/view
+
+## Sources
+
+TBD

curie_backend-0.1.0/README.md

@@ -0,0 +1,98 @@
+# Project Curie - Text2SQL
+
+
+
+## Project description
+* **Idea:** Use natural language prompts to generate SQL queries 
+* **Front-end: VS Code Plugin**
+  * Integration of trained LLM into useful plugin with quick insights 
+  * Establish connection with existing database 
+  * Use natural language to describe queries to this database, which are generated by fine-tuned LLM 
+  * Ability to run generated SQL code directly via plugin to see results immediately with no bothersome copying
+
+* ...
+
+### Simple example: 
+![simple_example.png](other/diagrams/simple_example.png)
+
+## Project structure
+
+```text
+evaluation/   Benchmark setups for Text-to-SQL evaluation
+data/         Local datasets and sample data
+docs/         Project documentation, presentations, and research notes
+experiments/  Ongoing fine-tuning experiments and run notes
+extension/    VS Code extension and demo database
+ml/           Training, fine-tuning, and inference code
+ml/models/    Model metadata, configs, and model cards
+```
+
+## Project team and responsibilities
+
+| Name              | E-Mail                   | Identifier | Responsibilities / tasks |
+|-------------------|--------------------------|------------|--------------------------|
+| Tamer Lewis       | tamer.lewis@tum.de       | go68maj    | TBD                      |
+| Antonia Gilles    | antonia.gilles@tum.de    | go34cih    | TBD                      |
+| Leon Pfänder      | leon.pfaender@tum.de     | ge27fob    | TBD                      |
+| Christoph Klein   | christoph.m.klein@tum.de | ge42low    | TBD                      |
+
+
+
+
+## Project setup
+
+The project uses Python 3.12 and a pinned ML stack from `requirements.txt`.
+
+```bash
+python3.12 -m venv .venv
+source .venv/bin/activate
+python -m pip install --upgrade pip
+python -m pip install -r requirements.txt
+```
+
+Verify that the installed dependencies are consistent:
+
+```bash
+python -m pip check
+```
+
+### Git LFS for model weights
+
+Git LFS stores large model binaries on the GitLab LFS server while Git tracks
+small pointer files. This repository uses LFS for `*.safetensors` files.
+
+Minimal setup:
+
+```bash
+brew install git-lfs
+git lfs install
+git lfs pull
+```
+
+Add and push model weights through the normal Git workflow:
+
+```bash
+git add .gitattributes ml/models/<model>/adapter_model.safetensors
+git commit -m "Add model adapter"
+git push
+```
+
+After cloning, Git LFS normally downloads tracked files automatically. If a model file is only an LFS pointer or is missing, run `git lfs pull`. Verify the available model files with `git lfs ls-files`.
+
+LFS does not change how the benchmark is invoked:
+
+```bash
+python3 evaluation/spider/run.py \
+  --split dev \
+  --model-dir ml/models/<model>
+```
+
+### Spider
+
+Download the Spider dataset and place it in the `data/` folder:
+
+https://drive.google.com/file/d/1403EGqzIDoHMdQF4c9Bkyl7dZLZ5Wt6J/view
+
+## Sources
+
+TBD

curie_backend-0.1.0/curie_backend.egg-info/PKG-INFO

@@ -0,0 +1,116 @@
+Metadata-Version: 2.4
+Name: curie-backend
+Version: 0.1.0
+Summary: Text2SQL Assistant Backend
+Classifier: Programming Language :: Python :: 3
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+Requires-Dist: fastapi[standard]>=0.110.0
+Requires-Dist: uvicorn>=0.28.0
+Requires-Dist: torch>=2.1.0
+Requires-Dist: transformers>=4.38.0
+Requires-Dist: peft>=0.9.0
+Requires-Dist: accelerate>=0.27.0
+Requires-Dist: SQLAlchemy>=2.0.0
+Requires-Dist: psycopg[binary]>=3.1.0
+Requires-Dist: huggingface-hub>=0.21.0
+Requires-Dist: python-dotenv>=1.0.0
+
+# Project Curie - Text2SQL
+
+
+
+## Project description
+* **Idea:** Use natural language prompts to generate SQL queries 
+* **Front-end: VS Code Plugin**
+  * Integration of trained LLM into useful plugin with quick insights 
+  * Establish connection with existing database 
+  * Use natural language to describe queries to this database, which are generated by fine-tuned LLM 
+  * Ability to run generated SQL code directly via plugin to see results immediately with no bothersome copying
+
+* ...
+
+### Simple example: 
+![simple_example.png](other/diagrams/simple_example.png)
+
+## Project structure
+
+```text
+evaluation/   Benchmark setups for Text-to-SQL evaluation
+data/         Local datasets and sample data
+docs/         Project documentation, presentations, and research notes
+experiments/  Ongoing fine-tuning experiments and run notes
+extension/    VS Code extension and demo database
+ml/           Training, fine-tuning, and inference code
+ml/models/    Model metadata, configs, and model cards
+```
+
+## Project team and responsibilities
+
+| Name              | E-Mail                   | Identifier | Responsibilities / tasks |
+|-------------------|--------------------------|------------|--------------------------|
+| Tamer Lewis       | tamer.lewis@tum.de       | go68maj    | TBD                      |
+| Antonia Gilles    | antonia.gilles@tum.de    | go34cih    | TBD                      |
+| Leon Pfänder      | leon.pfaender@tum.de     | ge27fob    | TBD                      |
+| Christoph Klein   | christoph.m.klein@tum.de | ge42low    | TBD                      |
+
+
+
+
+## Project setup
+
+The project uses Python 3.12 and a pinned ML stack from `requirements.txt`.
+
+```bash
+python3.12 -m venv .venv
+source .venv/bin/activate
+python -m pip install --upgrade pip
+python -m pip install -r requirements.txt
+```
+
+Verify that the installed dependencies are consistent:
+
+```bash
+python -m pip check
+```
+
+### Git LFS for model weights
+
+Git LFS stores large model binaries on the GitLab LFS server while Git tracks
+small pointer files. This repository uses LFS for `*.safetensors` files.
+
+Minimal setup:
+
+```bash
+brew install git-lfs
+git lfs install
+git lfs pull
+```
+
+Add and push model weights through the normal Git workflow:
+
+```bash
+git add .gitattributes ml/models/<model>/adapter_model.safetensors
+git commit -m "Add model adapter"
+git push
+```
+
+After cloning, Git LFS normally downloads tracked files automatically. If a model file is only an LFS pointer or is missing, run `git lfs pull`. Verify the available model files with `git lfs ls-files`.
+
+LFS does not change how the benchmark is invoked:
+
+```bash
+python3 evaluation/spider/run.py \
+  --split dev \
+  --model-dir ml/models/<model>
+```
+
+### Spider
+
+Download the Spider dataset and place it in the `data/` folder:
+
+https://drive.google.com/file/d/1403EGqzIDoHMdQF4c9Bkyl7dZLZ5Wt6J/view
+
+## Sources
+
+TBD

curie_backend-0.1.0/curie_backend.egg-info/SOURCES.txt

@@ -0,0 +1,24 @@
+README.md
+pyproject.toml
+curie_backend.egg-info/PKG-INFO
+curie_backend.egg-info/SOURCES.txt
+curie_backend.egg-info/dependency_links.txt
+curie_backend.egg-info/entry_points.txt
+curie_backend.egg-info/requires.txt
+curie_backend.egg-info/top_level.txt
+extension/backend/__init__.py
+extension/backend/__main__.py
+extension/backend/config.py
+extension/backend/main.py
+extension/backend/schemas.py
+extension/backend/routes/__init__.py
+extension/backend/routes/generation.py
+extension/backend/services/__init__.py
+extension/backend/services/database_service.py
+extension/backend/services/model_service.py
+ml/__init__.py
+ml/finetune.py
+ml/inference.py
+ml/load_datasets.py
+ml/model_loader.py
+ml/prompting.py

curie_backend-0.1.0/curie_backend.egg-info/dependency_links.txt

@@ -0,0 +1 @@
+

curie_backend-0.1.0/curie_backend.egg-info/entry_points.txt

@@ -0,0 +1,2 @@
+[console_scripts]
+curie-backend = extension.backend.__main__:main

curie_backend-0.1.0/curie_backend.egg-info/requires.txt

@@ -0,0 +1,10 @@
+fastapi[standard]>=0.110.0
+uvicorn>=0.28.0
+torch>=2.1.0
+transformers>=4.38.0
+peft>=0.9.0
+accelerate>=0.27.0
+SQLAlchemy>=2.0.0
+psycopg[binary]>=3.1.0
+huggingface-hub>=0.21.0
+python-dotenv>=1.0.0

curie_backend-0.1.0/curie_backend.egg-info/top_level.txt

@@ -0,0 +1,2 @@
+extension
+ml

curie_backend-0.1.0/extension/backend/__init__.py

@@ -0,0 +1 @@
+"""Backend package."""

curie_backend-0.1.0/extension/backend/__main__.py

@@ -0,0 +1,20 @@
+import argparse
+import uvicorn
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(description="Curie Backend Server")
+    parser.add_argument("--host", default="127.0.0.1", help="Host address to bind to")
+    parser.add_argument("--port", type=int, default=8000, help="Port to listen on")
+    args = parser.parse_args()
+
+    uvicorn.run(
+        "extension.backend.main:app",
+        host=args.host,
+        port=args.port,
+    )
+
+
+if __name__ == "__main__":
+    main()
+

curie_backend-0.1.0/extension/backend/config.py

@@ -0,0 +1,26 @@
+from dataclasses import dataclass
+from pathlib import Path
+
+from ml.inference import GenerationConfig
+from ml.model_loader import Device
+
+PROJECT_ROOT = Path(__file__).resolve().parents[2]
+CACHE_DIR = Path.home() / ".cache" / "curie"
+
+
+@dataclass(frozen=True)
+class Settings:
+    project_root: Path = PROJECT_ROOT
+    models_dir: Path = CACHE_DIR / "models"
+    model_dir: Path = CACHE_DIR / "models" / "llama-3.2-1b-spider-lora-3000ex"
+    prompt_path: Path = PROJECT_ROOT / "ml" / "prompts" / "test_prompt.txt"
+    device: Device = "auto"
+    generation: GenerationConfig = GenerationConfig(
+        max_new_tokens=128,
+        temperature=0.0,
+    )
+
+
+
+settings = Settings()
+

curie_backend-0.1.0/extension/backend/main.py

@@ -0,0 +1,25 @@
+from collections.abc import AsyncIterator
+from contextlib import asynccontextmanager
+
+from fastapi import FastAPI
+
+from extension.backend.routes.generation import router
+from extension.backend.services.model_service import ModelService
+
+
+@asynccontextmanager
+async def lifespan(app: FastAPI) -> AsyncIterator[None]:
+    service = ModelService()
+    service.load()
+
+    app.state.model_service = service
+    yield
+
+
+app = FastAPI(title="Curie Backend", lifespan=lifespan)
+app.include_router(router, prefix="/api")
+
+
+@app.get("/health")
+def health() -> dict[str, str]:
+    return {"status": "ok"}

curie_backend-0.1.0/extension/backend/routes/__init__.py

@@ -0,0 +1 @@
+"""Backend route package."""

curie_backend-0.1.0/extension/backend/routes/generation.py

@@ -0,0 +1,99 @@
+from fastapi import APIRouter, HTTPException, Request
+from extension.backend.schemas import (
+    ExecuteRequest,
+    ExecuteResponse,
+    GenerationRequest,
+    GenerationResponse,
+    SchemaRequest,
+    SchemaResponse,
+    AvailableModelResponse,
+    ModelsResponse,
+    SelectModelRequest,
+)
+from extension.backend.services.database_service import (
+    DatabaseConnectionError,
+    DatabaseService,
+    InvalidSqlError,
+    QueryExecutionError,
+)
+from extension.backend.services.model_service import ModelService
+
+router = APIRouter()
+
+
+@router.get("/models", response_model=ModelsResponse)
+def list_models(request: Request) -> ModelsResponse:
+    model_service: ModelService = request.app.state.model_service
+
+    return ModelsResponse(
+        models=[
+            AvailableModelResponse(
+                id=model.id,
+                name=model.name,
+                is_active=model.is_active,
+            )
+            for model in model_service.list_models()
+        ]
+    )
+
+
+@router.post("/models/select", response_model=AvailableModelResponse)
+def select_model(body: SelectModelRequest, request: Request) -> AvailableModelResponse:
+    model_service: ModelService = request.app.state.model_service
+
+    try:
+        model = model_service.select_model(body.model_id)
+    except ValueError as error:
+        raise HTTPException(status_code=404, detail=str(error)) from error
+
+    return AvailableModelResponse(
+        id=model.id,
+        name=model.name,
+        is_active=model.is_active,
+    )
+
+
+@router.post("/generate", response_model=GenerationResponse)
+def generate_sql(body: GenerationRequest, request: Request) -> GenerationResponse:
+    model_service: ModelService = request.app.state.model_service
+
+    if body.database is not None:
+        database_service = DatabaseService()
+        try:
+            schema = database_service.build_schema_text(body.database)
+        except DatabaseConnectionError as error:
+            raise HTTPException(status_code=400, detail=str(error)) from error
+
+        sql = model_service.generate_sql(body.question, schema)
+        return GenerationResponse(sql=sql)
+
+    if body.database_schema is None:
+        raise HTTPException(status_code=422, detail="Either schema or database must be provided")
+
+    sql = model_service.generate_sql(body.question, body.database_schema)
+
+    return GenerationResponse(sql=sql)
+
+
+@router.post("/execute", response_model=ExecuteResponse)
+def execute_sql(body: ExecuteRequest) -> ExecuteResponse:
+    database_service = DatabaseService()
+    try:
+        result = database_service.execute(body.sql, body.database, body.read_only)
+    except InvalidSqlError as error:
+        raise HTTPException(status_code=400, detail=str(error)) from error
+    except DatabaseConnectionError as error:
+        raise HTTPException(status_code=400, detail=str(error)) from error
+    except QueryExecutionError as error:
+        raise HTTPException(status_code=400, detail=str(error)) from error
+
+    return ExecuteResponse(**result)
+
+
+@router.post("/schema", response_model=SchemaResponse)
+def inspect_schema(body: SchemaRequest) -> SchemaResponse:
+    database_service = DatabaseService()
+    try:
+        return database_service.inspect_schema(body.database)
+    except DatabaseConnectionError as error:
+        raise HTTPException(status_code=400, detail=str(error)) from error

curie_backend-0.1.0/extension/backend/schemas.py

@@ -0,0 +1,83 @@
+from typing import Any, Literal
+
+from pydantic import BaseModel, ConfigDict, Field, model_validator
+
+
+class PostgreSQLDatabaseConfig(BaseModel):
+    type: Literal["postgresql"]
+    host: str = Field(min_length=1)
+    port: int = Field(default=5432, ge=1, le=65535)
+    database: str = Field(min_length=1)
+    username: str = Field(min_length=1)
+    password: str
+    sslmode: Literal["disable", "prefer", "require"] = "prefer"
+
+
+class GenerationRequest(BaseModel):
+    model_config = ConfigDict(populate_by_name=True)
+
+    question: str = Field(min_length=1)
+    database_schema: str | None = Field(default=None, alias="schema", min_length=1)
+    database: PostgreSQLDatabaseConfig | None = None
+
+    @model_validator(mode="after")
+    def require_schema_or_database(self) -> "GenerationRequest":
+        if self.database_schema is None and self.database is None:
+            raise ValueError("Either schema or database must be provided")
+        return self
+
+
+class GenerationResponse(BaseModel):
+    sql: str
+
+
+class DatabaseColumn(BaseModel):
+    name: str
+    type: str
+    primary_key: bool
+
+
+class DatabaseForeignKey(BaseModel):
+    columns: list[str]
+    referred_table: str
+    referred_columns: list[str]
+
+
+class DatabaseTable(BaseModel):
+    name: str
+    columns: list[DatabaseColumn]
+    foreign_keys: list[DatabaseForeignKey]
+
+
+class SchemaRequest(BaseModel):
+    database: PostgreSQLDatabaseConfig
+
+
+class SchemaResponse(BaseModel):
+    tables: list[DatabaseTable]
+
+
+class ExecuteRequest(BaseModel):
+    sql: str = Field(min_length=1)
+    database: PostgreSQLDatabaseConfig
+    read_only: bool = True
+
+
+class ExecuteResponse(BaseModel):
+    columns: list[str]
+    rows: list[dict[str, Any]]
+    row_count: int
+
+
+class AvailableModelResponse(BaseModel): 
+    id: str
+    name: str
+    is_active: bool
+
+
+class ModelsResponse(BaseModel): 
+    models: list[AvailableModelResponse]
+
+
+class SelectModelRequest(BaseModel):
+    model_id: str = Field(min_length=1)

curie_backend-0.1.0/extension/backend/services/__init__.py

@@ -0,0 +1 @@
+"""Backend service layer."""