DB - task_llm_assignments

abstract

Maps each AI task type (e.g., market_analysis, vision, execution_decision) to a specific LLM provider and model — allowing different tasks to use different models optimized for cost vs. capability.

Table Info

Property	Value
Table Name	`task_llm_assignments`
SQLAlchemy Model	`backend/db/models.py :: TaskLLMAssignment`
Pydantic Schema	`backend/api/routes/settings.py :: TaskAssignment / SaveAssignmentsRequest`
Migration File	`alembic/versions/`
TimescaleDB Hypertable	No
Partition Column	—

Columns

Column	SQLAlchemy Type	Nullable	Default	Description
`task`	`String(64)`	No	—	Task identifier (PRIMARY KEY)
`provider`	`String(32)`	No	—	LLM provider to use for this task
`model_name`	`String(128)`	No	—	Specific model ID (e.g., `claude-opus-4-7`)
`updated_at`	`DateTime(timezone=True)`	No	`datetime.now(UTC)`	Last update timestamp

Valid Task Values

From backend/api/routes/settings.py :: _VALID_TASKS:

Task	Description
`market_analysis`	Core market signal analysis
`vision`	Chart image interpretation
`execution_decision`	Final trade go/no-go decision
`maintenance_technical`	Technical indicator check for open trade
`maintenance_sentiment`	News/sentiment check for open trade
`maintenance_decision`	Final decision on open trade management
`news_analysis`	Financial news parsing and impact scoring
`post_trade_analysis`	Post-close trade review and learning
`indicator_agent`	Dedicated indicator analysis agent

Constraints & Indexes

Name	Type	Columns	Purpose
`pk_task_llm_assignments`	PRIMARY KEY	`task`	One assignment per task type

SQLAlchemy Model (reference snapshot)

class TaskLLMAssignment(Base):
    __tablename__ = "task_llm_assignments"
 
    task: Mapped[str] = mapped_column(String(64), primary_key=True)
    # 'market_analysis' | 'vision' | 'execution_decision' | ...
    provider: Mapped[str] = mapped_column(String(32), nullable=False)
    model_name: Mapped[str] = mapped_column(String(128), nullable=False)
    updated_at: Mapped[datetime] = mapped_column(
        DateTime(timezone=True),
        default=lambda: datetime.now(UTC),
        onupdate=lambda: datetime.now(UTC),
    )