feat: support document and collection summary

Author: iziang

Makefile

@@ -140,7 +140,7 @@ compose-logs:
 # Local development services
 .PHONY: run-backend run-frontend run-celery run-flower run-beat
 run-backend: migrate
-	uvicorn aperag.app:app --host 0.0.0.0 --reload --log-config scripts/uvicorn-log-config.yaml
+	uvicorn aperag.app:app --host 0.0.0.0 --log-config scripts/uvicorn-log-config.yaml
 
 run-celery:
 	celery -A config.celery worker -B -l INFO --pool=threads --concurrency=16

aperag/agent/rag_agent.py

@@ -28,8 +28,8 @@
 from mcp_agent.config import LoggerSettings, MCPServerSettings, MCPSettings, OpenAISettings, Settings
 from mcp_agent.workflows.llm.augmented_llm_openai import OpenAIAugmentedLLM
 
-os.environ["APERAG_API_KEY"] = "sk-test"
-os.environ["OPENAI_API_KEY"] = "sk-test"
+os.environ["APERAG_API_KEY"] = ""
+os.environ["OPENAI_API_KEY"] = ""
 os.environ["APERAG_URL"] = "http://localhost:8000/mcp/"
 os.environ["OPENAI_BASE_URL"] = "https://openrouter.ai/api/v1"
 os.environ["DEFAULT_MODEL"] = "gpt-4o-mini"
@@ -262,7 +262,7 @@ async def interactive_chat(self):
                             continue
 
                         print("🔍 Searching knowledge base...")
-                        response = await llm.generate_str(question)
+                        response = await llm.generate(question)
                         print(f"\n🤖 **Answer:**\n{response}")
 
                     except KeyboardInterrupt:

aperag/api/components/schemas/collection.yaml

@@ -148,6 +148,9 @@ collectionConfig:
     enable_knowledge_graph:
       type: boolean
       description: Whether to enable knowledge graph
+    enable_summary:
+      type: boolean
+      description: Whether to enable summary generation
     embedding:
       $ref: './model.yaml#/modelSpec'
     completion:
@@ -271,6 +274,9 @@ collection:
       type: string
     description:
       type: string
+    summary:
+      type: string
+      description: LLM-generated summary of the collection
     config:
       $ref: '#/collectionConfig'
     source:
@@ -281,6 +287,7 @@ collection:
         - ACTIVE
         - INACTIVE
         - DELETED
+        - SUMMARY_GENERATING
     created:
       type: string
       format: date-time

aperag/api/components/schemas/document.yaml

@@ -44,6 +44,16 @@ document:
         - DELETION_IN_PROGRESS
         - FAILED
         - SKIPPED
+    summary_index_status:
+      type: string
+      enum:
+        - PENDING
+        - CREATING
+        - ACTIVE
+        - DELETING
+        - DELETION_IN_PROGRESS
+        - FAILED
+        - SKIPPED
     vector_index_updated:
       type: string
       format: date-time
@@ -56,6 +66,13 @@ document:
       type: string
       format: date-time
       description: Graph index last updated time
+    summary_index_updated:
+      type: string
+      format: date-time
+      description: Summary index last updated time
+    summary:
+      type: string
+      description: Summary of the document
     config:
       type: string
     size:
@@ -101,6 +118,7 @@ rebuildIndexesRequest:
           - VECTOR
           - FULLTEXT
           - GRAPH
+          - SUMMARY
       description: Types of indexes to rebuild
       minItems: 1
   required:

aperag/api/openapi.yaml

@@ -43,6 +43,8 @@ paths:
     $ref: './paths/collections.yaml#/collections'
   /collections/{collection_id}:
     $ref: './paths/collections.yaml#/collection'
+  /collections/{collection_id}/summary/generate:
+    $ref: './paths/collections.yaml#/summary_generate'
   /collections/{collection_id}/documents:
     $ref: './paths/collections.yaml#/documents'
   /collections/{collection_id}/documents/{document_id}:

aperag/api/paths/collections.yaml

@@ -120,6 +120,41 @@ collection:
             schema:
               $ref: '../components/schemas/collection.yaml#/collection'
 
+summary_generate:
+  post:
+    summary: Generate collection summary
+    description: Trigger collection summary generation as background task
+    security:
+      - BearerAuth: []
+    parameters:
+      - name: collection_id
+        in: path
+        required: true
+        schema:
+          type: string
+    responses:
+      '200':
+        description: Collection summary generation started
+        content:
+          application/json:
+            schema:
+              type: object
+              properties:
+                collection_id:
+                  type: string
+                success:
+                  type: boolean
+                message:
+                  type: string
+                status:
+                  type: string
+      '404':
+        description: Collection not found
+        content:
+          application/json:
+            schema:
+              $ref: '../components/schemas/common.yaml#/failResponse'
+
 documents:
   get:
     summary: List documents

aperag/app.py

@@ -13,6 +13,7 @@
 # limitations under the License.
 
 import os
+
 from fastapi import FastAPI
 
 from aperag.exception_handlers import register_exception_handlers
@@ -65,6 +66,7 @@ async def health_check():
 # Only include test router in dev mode
 if os.environ.get("DEPLOYMENT_MODE") == "dev":
     from aperag.views.test import router as test_router
+
     app.include_router(test_router, prefix="/api/v1")
 
 # Mount the MCP server at /mcp path

aperag/config.py

@@ -176,8 +176,7 @@ def __init__(self, **kwargs):
         # CELERY_BROKER_URL
         if not self.celery_broker_url:
             self.celery_broker_url = (
-                f"redis://{self.redis_user}:{self.redis_password}"
-                f"@{self.redis_host}:{self.redis_port}/0"
+                f"redis://{self.redis_user}:{self.redis_password}@{self.redis_host}:{self.redis_port}/0"
             )
 
         # CELERY_RESULT_BACKEND
@@ -187,20 +186,16 @@ def __init__(self, **kwargs):
         # MEMORY_REDIS_URL
         if not self.memory_redis_url:
             self.memory_redis_url = (
-                f"redis://{self.redis_user}:{self.redis_password}"
-                f"@{self.redis_host}:{self.redis_port}/1"
+                f"redis://{self.redis_user}:{self.redis_password}@{self.redis_host}:{self.redis_port}/1"
             )
         # ES_HOST
         if not self.es_host:
             if self.es_user and self.es_password:
                 self.es_host = (
-                    f"{self.es_protocol}://{self.es_user}:{self.es_password}"
-                    f"@{self.es_host_name}:{self.es_port}"
+                    f"{self.es_protocol}://{self.es_user}:{self.es_password}@{self.es_host_name}:{self.es_port}"
                 )
             else:
-                self.es_host = (
-                    f"{self.es_protocol}://{self.es_host_name}:{self.es_port}"
-                )
+                self.es_host = f"{self.es_protocol}://{self.es_host_name}:{self.es_port}"
         # Object store config
         if self.object_store_type == "local":
             self.object_store_local_config = LocalObjectStoreConfig()

aperag/db/models.py

@@ -62,6 +62,7 @@ class CollectionStatus(str, Enum):
     INACTIVE = "INACTIVE"
     ACTIVE = "ACTIVE"
     DELETED = "DELETED"
+    SUMMARY_GENERATING = "SUMMARY_GENERATING"  # Currently generating collection summary
 
 
 class CollectionType(str, Enum):
@@ -82,6 +83,7 @@ class DocumentIndexType(str, Enum):
     VECTOR = "VECTOR"
     FULLTEXT = "FULLTEXT"
     GRAPH = "GRAPH"
+    SUMMARY = "SUMMARY"
 
 
 class DocumentIndexStatus(str, Enum):
@@ -168,7 +170,8 @@ class Collection(Base):
 
     id = Column(String(24), primary_key=True, default=lambda: "col" + random_id())
     title = Column(String(256), nullable=False)
-    description = Column(Text, nullable=True)
+    description = Column(Text, nullable=True)  # User-provided description
+    summary = Column(Text, nullable=True)  # LLM-generated summary
     user = Column(String(256), nullable=False, index=True)  # Add index for frequent queries
     status = Column(EnumColumn(CollectionStatus), nullable=False, index=True)  # Add index for status queries
     type = Column(EnumColumn(CollectionType), nullable=False)

aperag/index/base.py

@@ -24,6 +24,7 @@ class IndexType(Enum):
     VECTOR = "VECTOR"
     FULLTEXT = "FULLTEXT"
     GRAPH = "GRAPH"
+    SUMMARY = "SUMMARY"
 
 
 class IndexStatus(Enum):