remdb 0.3.7__py3-none-any.whl → 0.3.133__py3-none-any.whl

rem/api/routers/query.py

@@ -0,0 +1,360 @@
+"""
+REM Query API - Execute REM dialect or natural language queries.
+
+Endpoints:
+    POST /api/v1/query - Execute a REM query
+
+Modes:
+    - rem-dialect (default): Execute REM query syntax directly
+      Example: "LOOKUP sarah-chen", "SEARCH resources 'API design'", "TRAVERSE FROM doc-123 DEPTH 2"
+
+    - natural-language: Convert natural language to REM query via LLM agent
+      Example: "Find all documents by Sarah", "What meetings happened last week?"
+
+    - staged-plan: Execute a multi-stage query plan (query field is ignored)
+      Example: Execute a sequence of queries with context passing between stages
+      Status: TODO - signature only, implementation pending in RemService
+
+Model Selection:
+    Default model: openai:gpt-4.1 (widely available, good balance of speed/quality)
+
+    Recommended for speed: cerebras:qwen-3-32b
+    - Cerebras provides extremely fast inference (~1000 tokens/sec)
+    - Set CEREBRAS_API_KEY environment variable
+    - Pass model="cerebras:qwen-3-32b" in request
+
+Example:
+    # REM dialect (default)
+    curl -X POST http://localhost:8000/api/v1/query \\
+        -H "Content-Type: application/json" \\
+        -H "X-User-Id: user123" \\
+        -d '{"query": "LOOKUP sarah-chen"}'
+
+    # Natural language
+    curl -X POST http://localhost:8000/api/v1/query \\
+        -H "Content-Type: application/json" \\
+        -H "X-User-Id: user123" \\
+        -d '{"query": "Find all documents about API design", "mode": "natural-language"}'
+
+    # With Cerebras for speed
+    curl -X POST http://localhost:8000/api/v1/query \\
+        -H "Content-Type: application/json" \\
+        -H "X-User-Id: user123" \\
+        -d '{"query": "Who is Sarah?", "mode": "natural-language", "model": "cerebras:qwen-3-32b"}'
+
+    # Staged plan (TODO) - static query stages
+    curl -X POST http://localhost:8000/api/v1/query \\
+        -H "Content-Type: application/json" \\
+        -H "X-User-Id: user123" \\
+        -d '{"mode": "staged-plan", "plan": [
+            {"stage": 1, "query": "LOOKUP Sarah Chen", "name": "user"},
+            {"stage": 2, "query": "TRAVERSE FROM \"Sarah Chen\" DEPTH 2"}
+        ]}'
+
+    # Staged plan with LLM-driven dynamic stages
+    curl -X POST http://localhost:8000/api/v1/query \\
+        -H "Content-Type: application/json" \\
+        -H "X-User-Id: user123" \\
+        -d '{"mode": "staged-plan", "plan": [
+            {"stage": 1, "query": "LOOKUP Sarah Chen", "name": "user"},
+            {"stage": 2, "intent": "find her team members", "depends_on": ["user"]}
+        ]}'
+
+    # Plan continuation - pass previous_results to resume a multi-turn plan
+    # Turn 1: Execute stage 1, get back stage_results
+    # Turn 2: Continue with stage 2, passing previous results
+    curl -X POST http://localhost:8000/api/v1/query \\
+        -H "Content-Type: application/json" \\
+        -H "X-User-Id: user123" \\
+        -d '{
+            "mode": "staged-plan",
+            "plan": [
+                {"stage": 1, "query": "LOOKUP Sarah Chen", "name": "user"},
+                {"stage": 2, "intent": "find her team members", "depends_on": ["user"]}
+            ],
+            "previous_results": [
+                {"stage": 1, "name": "user", "query_executed": "LOOKUP Sarah Chen", "results": [...], "count": 1}
+            ],
+            "resume_from_stage": 2
+        }'
+"""
+
+from enum import Enum
+from typing import Any
+
+from fastapi import APIRouter, Header, HTTPException
+from loguru import logger
+from pydantic import BaseModel, Field
+
+from ...services.postgres import get_postgres_service
+from ...services.rem.service import RemService
+from ...services.rem.parser import RemQueryParser
+from ...models.core import RemQuery
+from ...settings import settings
+
+router = APIRouter(prefix="/api/v1", tags=["query"])
+
+
+class QueryMode(str, Enum):
+    """Query execution mode."""
+    REM_DIALECT = "rem-dialect"
+    NATURAL_LANGUAGE = "natural-language"
+    STAGED_PLAN = "staged-plan"
+
+
+class StagedPlanResult(BaseModel):
+    """Result from a completed stage - used for plan continuation."""
+
+    stage: int = Field(..., description="Stage number that produced this result")
+    name: str | None = Field(default=None, description="Stage name for referencing")
+    query_executed: str = Field(..., description="The REM query that was executed")
+    results: list[dict[str, Any]] = Field(default_factory=list, description="Query results")
+    count: int = Field(default=0, description="Number of results")
+
+
+class QueryPlanStage(BaseModel):
+    """A single stage in a multi-stage query plan.
+
+    Each stage can be either:
+    1. A static REM dialect query (query field set)
+    2. A dynamic query built by LLM from intent + previous results (intent field set)
+
+    The LLM interprets the intent along with previous stage results to construct
+    the appropriate REM query at runtime.
+    """
+
+    stage: int = Field(..., description="Stage number (1-indexed, executed in order)")
+    query: str | None = Field(
+        default=None,
+        description="Static REM dialect query (mutually exclusive with intent)",
+    )
+    intent: str | None = Field(
+        default=None,
+        description="Natural language intent - LLM builds query from this + previous results",
+    )
+    name: str | None = Field(default=None, description="Optional name for referencing results")
+    depends_on: list[str] | None = Field(
+        default=None,
+        description="Names of previous stages whose results are passed as context to LLM",
+    )
+
+
+class QueryRequest(BaseModel):
+    """Request body for REM query execution."""
+
+    query: str | None = Field(
+        default=None,
+        description="Query string - either REM dialect syntax or natural language. Required for rem-dialect and natural-language modes.",
+        examples=[
+            "LOOKUP sarah-chen",
+            "SEARCH resources 'API design' LIMIT 10",
+            "Find all documents by Sarah",
+        ],
+    )
+
+    mode: QueryMode = Field(
+        default=QueryMode.REM_DIALECT,
+        description="Query mode: 'rem-dialect' (default), 'natural-language', or 'staged-plan'",
+    )
+
+    model: str = Field(
+        default="openai:gpt-4.1",
+        description=(
+            "LLM model for natural-language mode. "
+            "Default: openai:gpt-4.1. "
+            "Recommended for speed: cerebras:qwen-3-32b (requires CEREBRAS_API_KEY)"
+        ),
+    )
+
+    plan_only: bool = Field(
+        default=False,
+        description="If true with natural-language mode, return generated query without executing",
+    )
+
+    plan: list[QueryPlanStage] | None = Field(
+        default=None,
+        description="Multi-stage query plan for staged-plan mode. Each stage executes in order.",
+    )
+
+    previous_results: list[StagedPlanResult] | None = Field(
+        default=None,
+        description=(
+            "Results from previous turns for plan continuation. "
+            "Pass this back from the response's stage_results to continue a multi-turn plan."
+        ),
+    )
+
+    resume_from_stage: int | None = Field(
+        default=None,
+        description="Stage number to resume from (1-indexed). Stages before this are skipped.",
+    )
+
+
+class QueryResponse(BaseModel):
+    """Response from REM query execution."""
+
+    query_type: str = Field(..., description="Type of query executed (LOOKUP, SEARCH, FUZZY, SQL, TRAVERSE)")
+    query: str = Field(..., description="The query that was executed (original or generated)")
+    results: list[dict[str, Any]] = Field(default_factory=list, description="Query results")
+    count: int = Field(..., description="Number of results")
+
+    # Natural language mode fields
+    mode: QueryMode = Field(..., description="Query mode used")
+    generated_query: str | None = Field(default=None, description="Generated REM query (natural-language mode only)")
+    confidence: float | None = Field(default=None, description="Confidence score (natural-language mode only)")
+    reasoning: str | None = Field(default=None, description="Query reasoning (natural-language mode only)")
+    warning: str | None = Field(default=None, description="Warning message if any")
+    plan_only: bool = Field(default=False, description="If true, query was not executed (plan mode)")
+
+    # Staged plan mode fields
+    stage_results: list[dict[str, Any]] | None = Field(
+        default=None,
+        description="Results from each stage (staged-plan mode only)",
+    )
+
+
+@router.post("/query", response_model=QueryResponse)
+async def execute_query(
+    request: QueryRequest,
+    x_user_id: str | None = Header(default=None, description="User ID for query isolation (optional, uses default if not provided)"),
+) -> QueryResponse:
+    """
+    Execute a REM query.
+
+    Supports three modes:
+
+    **rem-dialect** (default): Execute REM query syntax directly.
+    - LOOKUP "entity-key" - O(1) key-value lookup
+    - FUZZY "text" THRESHOLD 0.3 - Fuzzy text matching
+    - SEARCH table "semantic query" LIMIT 10 - Vector similarity search
+    - TRAVERSE FROM "entity" TYPE "rel" DEPTH 2 - Graph traversal
+    - SQL SELECT * FROM table WHERE ... - Direct SQL (SELECT only)
+
+    **natural-language**: Convert question to REM query via LLM.
+    - Uses REM Query Agent to parse intent
+    - Auto-executes if confidence >= 0.7
+    - Returns warning for low-confidence queries
+
+    **staged-plan**: Execute a multi-stage query plan.
+    - Pass plan=[{stage: 1, query: "...", name: "..."}, ...] instead of query
+    - Stages execute in order with context passing between them
+    - TODO: Implementation pending in RemService
+
+    **Model Selection**:
+    - Default: openai:gpt-4.1 (reliable, widely available)
+    - Speed: cerebras:qwen-3-32b (requires CEREBRAS_API_KEY)
+
+    Returns:
+        QueryResponse with results and metadata
+    """
+    if not settings.postgres.enabled:
+        raise HTTPException(
+            status_code=503,
+            detail="Database not configured. Set POSTGRES__ENABLED=true",
+        )
+
+    try:
+        # Get database service and ensure connected
+        db = get_postgres_service()
+        if db is None:
+            raise HTTPException(status_code=503, detail="Database service unavailable")
+
+        # Connect if not already connected
+        if db.pool is None:
+            await db.connect()
+
+        rem_service = RemService(db)
+
+        # Use effective_user_id from settings if not provided
+        effective_user_id = x_user_id or settings.test.effective_user_id
+
+        if request.mode == QueryMode.STAGED_PLAN:
+            # Staged plan mode - execute multi-stage query plan
+            # TODO: Implementation pending in RemService.execute_staged_plan()
+            if not request.plan:
+                raise HTTPException(
+                    status_code=400,
+                    detail="staged-plan mode requires 'plan' field with list of QueryPlanStage",
+                )
+
+            logger.info(f"Staged plan query: {len(request.plan)} stages")
+
+            # TODO: Call rem_service.execute_staged_plan(request.plan, x_user_id)
+            # For now, return a 501 Not Implemented
+            raise HTTPException(
+                status_code=501,
+                detail="staged-plan mode not yet implemented. See RemService TODO.",
+            )
+
+        elif request.mode == QueryMode.NATURAL_LANGUAGE:
+            # Natural language mode - use agent to convert
+            if not request.query:
+                raise HTTPException(
+                    status_code=400,
+                    detail="natural-language mode requires 'query' field",
+                )
+
+            logger.info(f"Natural language query: {request.query[:100]}... (model={request.model})")
+
+            result = await rem_service.ask_rem(
+                natural_query=request.query,
+                tenant_id=effective_user_id,
+                llm_model=request.model,
+                plan_mode=request.plan_only,
+            )
+
+            # Build response
+            response = QueryResponse(
+                query_type=result.get("results", {}).get("query_type", "UNKNOWN"),
+                query=request.query,
+                results=result.get("results", {}).get("results", []),
+                count=result.get("results", {}).get("count", 0),
+                mode=QueryMode.NATURAL_LANGUAGE,
+                generated_query=result.get("query"),
+                confidence=result.get("confidence"),
+                reasoning=result.get("reasoning"),
+                warning=result.get("warning"),
+                plan_only=result.get("plan_mode", False),
+            )
+
+            return response
+
+        else:
+            # REM dialect mode - parse and execute directly
+            if not request.query:
+                raise HTTPException(
+                    status_code=400,
+                    detail="rem-dialect mode requires 'query' field",
+                )
+
+            logger.info(f"REM dialect query: {request.query[:100]}...")
+
+            parser = RemQueryParser()
+            query_type, parameters = parser.parse(request.query)
+
+            # Create and execute RemQuery
+            rem_query = RemQuery.model_validate({
+                "query_type": query_type,
+                "parameters": parameters,
+                "user_id": effective_user_id,
+            })
+
+            result = await rem_service.execute_query(rem_query)
+
+            return QueryResponse(
+                query_type=result["query_type"],
+                query=request.query,
+                results=result.get("results", []),
+                count=result.get("count", 0),
+                mode=QueryMode.REM_DIALECT,
+            )
+
+    except HTTPException:
+        # Re-raise HTTPExceptions (400, 501, etc.) without wrapping
+        raise
+    except ValueError as e:
+        # Parse errors
+        raise HTTPException(status_code=400, detail=str(e))
+    except Exception as e:
+        logger.exception(f"Query execution failed: {e}")
+        raise HTTPException(status_code=500, detail=f"Query execution failed: {str(e)}")