Feature/llm chain

backend/app/alembic/versions/042_add_llm_call_table.py

@@ -0,0 +1,185 @@
+"""add_llm_call_table
+
+Revision ID: 042
+Revises: 041
+Create Date: 2026-01-26 15:20:23.873332
+
+"""
+from alembic import op
+import sqlalchemy as sa
+import sqlmodel.sql.sqltypes
+from sqlalchemy.dialects import postgresql
+
+# revision identifiers, used by Alembic.
+revision = "042"
+down_revision = "041"
+branch_labels = None
+depends_on = None
+
+
+def upgrade():
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.create_table(
+        "llm_call",
+        sa.Column(
+            "id",
+            sa.Uuid(),
+            nullable=False,
+            comment="Unique identifier for the LLM call record",
+        ),
+        sa.Column(
+            "job_id",
+            sa.Uuid(),
+            nullable=False,
+            comment="Reference to the parent job (status tracked in job table)",
+        ),
+        sa.Column(
+            "project_id",
+            sa.Integer(),
+            nullable=False,
+            comment="Reference to the project this LLM call belongs to",
+        ),
+        sa.Column(
+            "organization_id",
+            sa.Integer(),
+            nullable=False,
+            comment="Reference to the organization this LLM call belongs to",
+        ),
+        sa.Column(
+            "input",
+            sqlmodel.sql.sqltypes.AutoString(),
+            nullable=False,
+            comment="User input - text string, binary data, or file path for multimodal",
+        ),
+        sa.Column(
+            "input_type",
+            sa.String(),
+            nullable=False,
+            comment="Input type: text, audio, image",
+        ),
+        sa.Column(
+            "output_type",
+            sa.String(),
+            nullable=True,
+            comment="Expected output type: text, audio, image",
+        ),
+        sa.Column(
+            "provider",
+            sa.String(),
+            nullable=False,
+            comment="AI provider: openai, google, anthropic",
+        ),
+        sa.Column(
+            "model",
+            sqlmodel.sql.sqltypes.AutoString(),
+            nullable=False,
+            comment="Specific model used e.g. 'gpt-4o', 'gemini-2.5-pro'",
+        ),
+        sa.Column(
+            "provider_response_id",
+            sqlmodel.sql.sqltypes.AutoString(),
+            nullable=True,
+            comment="Original response ID from the provider (e.g., OpenAI's response ID)",
+        ),
+        sa.Column(
+            "content",
+            postgresql.JSONB(astext_type=sa.Text()),
+            nullable=True,
+            comment="Response content: {text: '...'}, {audio_bytes: '...'}, or {image: '...'}",
+        ),
+        sa.Column(
+            "usage",
+            postgresql.JSONB(astext_type=sa.Text()),
+            nullable=True,
+            comment="Token usage: {input_tokens, output_tokens, reasoning_tokens}",
+        ),
+        sa.Column(
+            "conversation_id",
+            sqlmodel.sql.sqltypes.AutoString(),
+            nullable=True,
+            comment="Identifier linking this response to its conversation thread",
+        ),
+        sa.Column(
+            "auto_create",
+            sa.Boolean(),
+            nullable=True,
+            comment="Whether to auto-create conversation if conversation_id doesn't exist (OpenAI specific)",
+        ),
+        sa.Column(
+            "config",
+            postgresql.JSONB(astext_type=sa.Text()),
+            nullable=True,
+            comment="Configuration: {config_id, config_version} for stored config OR {config_blob} for ad-hoc config",
+        ),
+        sa.Column(
+            "created_at",
+            sa.DateTime(),
+            nullable=False,
+            comment="Timestamp when the LLM call was created",
+        ),
+        sa.Column(
+            "updated_at",
+            sa.DateTime(),
+            nullable=False,
+            comment="Timestamp when the LLM call was last updated",
+        ),
+        sa.Column(
+            "deleted_at",
+            sa.DateTime(),
+            nullable=True,
+            comment="Timestamp when the record was soft-deleted",
+        ),
+        sa.ForeignKeyConstraint(["job_id"], ["job.id"], ondelete="CASCADE"),
+        sa.ForeignKeyConstraint(
+            ["organization_id"], ["organization.id"], ondelete="CASCADE"
+        ),
+        sa.ForeignKeyConstraint(["project_id"], ["project.id"], ondelete="CASCADE"),
+        sa.PrimaryKeyConstraint("id"),
+    )
+    op.create_index(
+        "idx_llm_call_conversation_id",
+        "llm_call",
+        ["conversation_id"],
+        unique=False,
+        postgresql_where=sa.text("conversation_id IS NOT NULL AND deleted_at IS NULL"),
+    )
+    op.create_index(
+        "idx_llm_call_job_id",
+        "llm_call",
+        ["job_id"],
+        unique=False,
+        postgresql_where=sa.text("deleted_at IS NULL"),
+    )
+    op.alter_column(
+        "collection",
+        "llm_service_name",
+        existing_type=sa.VARCHAR(),
+        comment="Name of the LLM service",
+        existing_comment="Name of the LLM service provider",
+        existing_nullable=False,
+    )
+    # ### end Alembic commands ###
+
+
+def downgrade():
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.alter_column(
+        "collection",
+        "llm_service_name",
+        existing_type=sa.VARCHAR(),
+        comment="Name of the LLM service provider",
+        existing_comment="Name of the LLM service",
+        existing_nullable=False,
+    )
+    op.drop_index(
+        "idx_llm_call_job_id",
+        table_name="llm_call",
+        postgresql_where=sa.text("deleted_at IS NULL"),
+    )
+    op.drop_index(
+        "idx_llm_call_conversation_id",
+        table_name="llm_call",
+        postgresql_where=sa.text("conversation_id IS NOT NULL AND deleted_at IS NULL"),
+    )
+    op.drop_table("llm_call")
+    # ### end Alembic commands ###

backend/app/alembic/versions/042_extend_collection_table_for_provider_.py → backend/app/alembic/versions/043_extend_collection_table_for_provider_.py

@@ -1,7 +1,7 @@
 """extend collection table for provider agnostic support
 
-Revision ID: 042
-Revises: 041
+Revision ID: 043
+Revises: 042
 Create Date: 2026-01-15 16:53:19.495583
 
 """
@@ -12,8 +12,8 @@
 
 
 # revision identifiers, used by Alembic.
-revision = "042"
-down_revision = "041"
+revision = "043"
+down_revision = "042"
 branch_labels = None
 depends_on = None
 

backend/app/alembic/versions/044_remove_enum_checks_llm_call_provider.py

@@ -0,0 +1,43 @@
+"""remove:enum checks llm_call provider
+
+Revision ID: 044
+Revises: 043
+Create Date: 2026-01-30 11:22:45.165543
+
+"""
+from alembic import op
+import sqlalchemy as sa
+import sqlmodel.sql.sqltypes
+
+
+# revision identifiers, used by Alembic.
+revision = "044"
+down_revision = "043"
+branch_labels = None
+depends_on = None
+
+
+def upgrade():
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.alter_column(
+        "llm_call",
+        "provider",
+        existing_type=sa.VARCHAR(),
+        comment="AI provider as sent by user (e.g openai, -native, google)",
+        existing_comment="AI provider: openai, google, anthropic",
+        existing_nullable=False,
+    )
+    # ### end Alembic commands ###
+
+
+def downgrade():
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.alter_column(
+        "llm_call",
+        "provider",
+        existing_type=sa.VARCHAR(),
+        comment="AI provider: openai, google, anthropic",
+        existing_comment="AI provider as sent by user (e.g openai, -native, google)",
+        existing_nullable=False,
+    )
+    # ### end Alembic commands ###

backend/app/api/routes/config/version.py

@@ -4,7 +4,7 @@
 from app.api.deps import SessionDep, AuthContextDep
 from app.crud.config import ConfigCrud, ConfigVersionCrud
 from app.models import (
-    ConfigVersionCreate,
+    ConfigVersionCreatePartial,
     ConfigVersionPublic,
     Message,
     ConfigVersionItems,
@@ -24,18 +24,21 @@
 )
 def create_version(
     config_id: UUID,
-    version_create: ConfigVersionCreate,
+    version_create: ConfigVersionCreatePartial,
     current_user: AuthContextDep,
     session: SessionDep,
 ):
     """
     Create a new version for an existing configuration.
-    The version number is automatically incremented.
+
+    Only include the fields you want to update in config_blob.
+    Provider, model, and params can be changed.
+    Type is inherited from existing config and cannot be changed.
     """
     version_crud = ConfigVersionCrud(
         session=session, project_id=current_user.project_.id, config_id=config_id
     )
-    version = version_crud.create_or_raise(version_create=version_create)
+    version = version_crud.create_from_partial_or_raise(version_create=version_create)
 
     return APIResponse.success_response(
         data=ConfigVersionPublic(**version.model_dump()),

backend/app/celery/beat.py

@@ -1,6 +1,7 @@
 """
 Celery beat scheduler for cron jobs.
 """
+
 import logging
 from celery import Celery
 from app.celery.celery_app import celery_app

backend/app/celery/utils.py

@@ -2,6 +2,7 @@
 Utility functions for easy Celery integration across the application.
 Business logic modules can use these functions without knowing Celery internals.
 """
+
 import logging
 from typing import Any, Dict, Optional
 from celery.result import AsyncResult

backend/app/celery/worker.py

@@ -1,6 +1,7 @@
 """
 Celery worker management script.
 """
+
 import logging
 import multiprocessing
 from celery.bin import worker

backend/app/cli/bench/commands.py

@@ -210,7 +210,7 @@ def send_benchmark_request(
         )
     else:
         typer.echo(response.text)
-        typer.echo(f"[{i+1}/{total}] FAILED - Status: {response.status_code}")
+        typer.echo(f"[{i + 1}/{total}] FAILED - Status: {response.status_code}")
         raise Exception(f"Request failed with status code {response.status_code}")
 
 

backend/app/core/providers.py

@@ -12,6 +12,7 @@ class Provider(str, Enum):
     OPENAI = "openai"
     AWS = "aws"
     LANGFUSE = "langfuse"
+    GOOGLE = "google"
 
 
 @dataclass
@@ -30,21 +31,11 @@ class ProviderConfig:
     Provider.LANGFUSE: ProviderConfig(
         required_fields=["secret_key", "public_key", "host"]
     ),
+    Provider.GOOGLE: ProviderConfig(required_fields=["api_key"]),
 }
 
 
 def validate_provider(provider: str) -> Provider:
-    """Validate that the provider name is supported and return the Provider enum.
-
-    Args:
-        provider: The provider name to validate
-
-    Returns:
-        Provider: The validated provider enum
-
-    Raises:
-        ValueError: If the provider is not supported
-    """
     try:
         return Provider(provider.lower())
     except ValueError: