prisma-web3-py 1.0.42

Python/SQLAlchemy async implementation of Prisma Web3 database models

Prisma Web3 Python

Async PostgreSQL ORM layer for HawkFi and downstream bots, built on SQLAlchemy 2.x async. Models mirror the Prisma schema (managed in the JS project) and expose repositories plus 2.x-style session helpers.

Contents

• Tables & Repositories
• Session & Config
• Install
• Usage
• Testing
• CI / GitHub Actions
• Cleanup
• Best Practices

Tables & Repositories

Model (table)	Repository	Purpose / notable fields
Token ("Token")	TokenRepository	Chain-normalized token catalog; search/fuzzy/alias; batch upsert (chain, token_address, symbol, aliases, platforms).
Signal ("Signal")	SignalRepository	Token alerts; source, signal_type, occurrence_count, last_occurrence.
PreSignal ("PreSignal")	PreSignalRepository	Early signals; heat scores (channel_calls, multi_signals, kol_discussions).
CryptoNews ("CryptoNews")	CryptoNewsRepository	News ingestion with unique source_link; JSONB fields matched_currencies, entity_list.
AIAnalysisResult ("AIAnalysisResult")	AIAnalysisRepository	Unified analysis (Twitter/News/etc.); JSONB heavy (tokens, key_points, final_signal, trading_recommendation, event_struct); JSONB search/stats helpers.
EventImpacts ("EventImpacts")	EventImpactRepository	Post-event pricing snapshots (price_t0/5m/15m/1h/4h, deltas, meta).
EventLabels ("EventLabels")	EventLabelsRepository	Label catalog for events.

Utilities: ChainConfig (chain normalization), datetime helpers, TokenImporter.

Session & Config

• URL: DATABASE_URL (or TEST_DATABASE_URL in tests). postgres:///postgresql:// auto-normalized to postgresql+asyncpg://.
• Pools: DB_POOL_SIZE, DB_POOL_OVERFLOW, DB_POOL_TIMEOUT.
• Engine/Session helpers (SQLAlchemy 2.x async):

from prisma_web3_py import session_scope, configure_engine, dispose_engine

# Optionally override URL (e.g., tests)
configure_engine("postgresql+asyncpg://user:pass@localhost:5432/prisma_web3")

async def main():
    async with session_scope() as session:  # commits on success, rollback on error
        ...
    async with session_scope(readonly=True) as session:  # read-only, no commit
        ...

Install

# editable dev install
pip install -e python[dev]

# or published package
pip install prisma-web3-py

Usage

from prisma_web3_py import session_scope, TokenRepository

repo = TokenRepository()

async def create_token():
    async with session_scope() as session:
        await repo.upsert_token(session, {"chain": "eth", "token_address": "0x...", "symbol": "UNI"})

async def search():
    async with session_scope(readonly=True) as session:
        tokens = await repo.search_by_symbol(session, "UNI")

Testing

• Unit-only (no DB required):

cd python
pytest -q tests/test_ai_analysis_utils.py tests/test_config_and_datetime.py tests/test_base_repository_unit.py

• Full + integration (requires Postgres, runs against the GH Actions Postgres service by default):

export DATABASE_URL=postgresql+asyncpg://user:pass@localhost:5432/prisma_web3_test
export RUN_DB_TESTS=1
cd python
pytest        # integration tests will run

• Schema setup: call python - <<'PY' to run init_db() (fallback create_all) if Prisma migrations aren’t applied yet. The CI workflow auto-runs init_db() against the Postgres service when PRISMA_DATABASE_URL is not provided.
• Cleanup after local/integration runs: python scripts/cleanup_test_data.py removes fake-llm / e2e analysis rows, test tokens/signals, and dangling EventImpacts/Labels.

CI / GitHub Actions

prisma-web3/.github/workflows/publish.yml runs unit + integration tests before publish.

• Default: uses the built-in Postgres service (postgresql+asyncpg://prisma:prisma@localhost:5432/prisma_web3_test) with RUN_DB_TESTS=1.
• Override with secrets (recommended for staging/prod-like DBs): set PRISMA_DATABASE_URL (asyncpg URL) and optionally PRISMA_RUN_DB_TESTS=1 under Settings → Secrets and variables → Actions.
• Schema bootstrap: when running against the service, the workflow calls configure_engine(...); init_db() to create tables; when pointing at an external DB it assumes Prisma migrations are pre-applied.

Cleanup

python/scripts/cleanup_test_data.py deletes test tokens/signals/pre-signals plus fake-llm/e2e AIAnalysisResult rows and dangling EventImpacts/Labels using session_scope.

Best Practices

• Use session_scope/get_db (alias) for unit-of-work; avoid manual commits in callers.
• Keep repositories thin; heavy normalization belongs in services. AIAnalysis repo currently mixes normalization with persistence—plan to extract to service layer.
• JSONB/pg_trgm features are Postgres-specific; document for consumers.
• Prefer ORM select().scalars(), insert().on_conflict_do_update(); avoid relying on rowcount to detect changes.
• Production deployments should run Prisma migrations; init_db() is a fallback for tests/local only.