Merge remote-tracking branch 'origin/main' into feat/upload-file-sandbox

Author: leehuwuj

.changeset/chilled-zebras-taste.md

@@ -0,0 +1,5 @@
+---
+"create-llama": patch
+---
+
+docs: chroma env variables

.changeset/great-students-appear.md

@@ -19,7 +19,7 @@ jobs:
         python-version: ["3.11"]
         os: [macos-latest, windows-latest, ubuntu-22.04]
         frameworks: ["fastapi"]
-        datasources: ["--no-files", "--example-file"]
+        datasources: ["--no-files", "--example-file", "--llamacloud"]
     defaults:
       run:
         shell: bash

.changeset/hungry-chairs-tie.md

@@ -17,6 +17,9 @@ jobs:
 
       - uses: pnpm/action-setup@v3
 
+      - name: Install uv
+        uses: astral-sh/setup-uv@v3
+
       - name: Setup Node.js
         uses: actions/setup-node@v4
         with:

.github/workflows/e2e.yml

@@ -1,5 +1,36 @@
 # create-llama
 
+## 0.2.19
+
+### Patch Changes
+
+- 3d41488: feat: use selected llamacloud for multiagent
+
+## 0.2.18
+
+### Patch Changes
+
+- 75e1f61: Fix cannot query public document from llamacloud
+- 88220f1: fix workflow doesn't stop when user presses stop generation button
+- 75e1f61: Fix typescript templates cannot upload file to llamacloud
+- 88220f1: Bump [email protected]
+
+## 0.2.17
+
+### Patch Changes
+
+- cd3fcd0: bump: use LlamaIndexTS 0.6.18
+- 6335de1: Fix using LlamaCloud selector does not use the configured values in the environment (Python)
+
+## 0.2.16
+
+### Patch Changes
+
+- 0e78ba4: Fix: programmatically ensure index for LlamaCloud
+- 0e78ba4: Fix .env not loaded on poetry run generate
+- 7f4ac22: Don't need to run generate script for LlamaCloud
+- 5263bde: Use selected LlamaCloud index in multi-agent template
+
 ## 0.2.15
 
 ### Patch Changes

.github/workflows/release.yml

@@ -27,6 +27,13 @@ const userMessage =
   dataSource !== "--no-files" ? "Physical standard for letters" : "Hello";
 
 test.describe(`Test streaming template ${templateFramework} ${dataSource} ${templateUI} ${appType} ${templatePostInstallAction}`, async () => {
+  const isNode18 = process.version.startsWith("v18");
+  const isLlamaCloud = dataSource === "--llamacloud";
+  // llamacloud is using File API which is not supported on node 18
+  if (isNode18 && isLlamaCloud) {
+    test.skip(true, "Skipping tests for Node 18 and LlamaCloud data source");
+  }
+
   let port: number;
   let externalPort: number;
   let cwd: string;

CHANGELOG.md

@@ -65,7 +65,7 @@ const getVectorDBEnvs = (
         {
           name: "PG_CONNECTION_STRING",
           description:
-            "For generating a connection URI, see https://docs.timescale.com/use-timescale/latest/services/create-a-service\nThe PostgreSQL connection string.",
+            "For generating a connection URI, see https://supabase.com/vector\nThe PostgreSQL connection string.",
         },
       ];
 
@@ -182,11 +182,11 @@ const getVectorDBEnvs = (
         },
         {
           name: "CHROMA_HOST",
-          description: "The API endpoint for your Chroma database",
+          description: "The hostname for your Chroma database. Eg: localhost",
         },
         {
           name: "CHROMA_PORT",
-          description: "The port for your Chroma database",
+          description: "The port for your Chroma database. Eg: 8000",
         },
       ];
       // TS Version doesn't support config local storage path

e2e/shared/streaming_template.spec.ts

@@ -1,6 +1,6 @@
 {
   "name": "create-llama",
-  "version": "0.2.15",
+  "version": "0.2.19",
   "description": "Create LlamaIndex-powered apps with one command",
   "keywords": [
     "rag",

helpers/env-variables.ts

@@ -1,6 +1,6 @@
 import {
+  BaseChatEngine,
   BaseToolWithCall,
-  ChatEngine,
   OpenAIAgent,
   QueryEngineTool,
 } from "llamaindex";
@@ -45,7 +45,7 @@ export async function createChatEngine(documentIds?: string[], params?: any) {
   const agent = new OpenAIAgent({
     tools,
     systemPrompt: process.env.SYSTEM_PROMPT,
-  }) as unknown as ChatEngine;
+  }) as unknown as BaseChatEngine;
 
   return agent;
 }

package.json

@@ -16,14 +16,26 @@ export async function uploadDocument(
     // trigger LlamaCloudIndex API to upload the file and run the pipeline
     const projectId = await index.getProjectId();
     const pipelineId = await index.getPipelineId();
-    return [
-      await LLamaCloudFileService.addFileToPipeline(
-        projectId,
-        pipelineId,
-        new File([fileBuffer], filename, { type: mimeType }),
-        { private: "true" },
-      ),
-    ];
+    try {
+      return [
+        await LLamaCloudFileService.addFileToPipeline(
+          projectId,
+          pipelineId,
+          new File([fileBuffer], filename, { type: mimeType }),
+          { private: "true" },
+        ),
+      ];
+    } catch (error) {
+      if (
+        error instanceof ReferenceError &&
+        error.message.includes("File is not defined")
+      ) {
+        throw new Error(
+          "File class is not supported in the current Node.js version. Please use Node.js 20 or higher.",
+        );
+      }
+      throw error;
+    }
   }
 
   // run the pipeline for other vector store indexes

templates/components/engines/typescript/agent/chat.ts

@@ -172,3 +172,26 @@ function getValidAnnotation(annotation: JSONValue): Annotation {
   }
   return { type: annotation.type, data: annotation.data };
 }
+
+// validate and get all annotations of a specific type or role from the frontend messages
+export function getAnnotations<
+  T extends Annotation["data"] = Annotation["data"],
+>(
+  messages: Message[],
+  options?: {
+    role?: Message["role"]; // message role
+    type?: Annotation["type"]; // annotation type
+  },
+): {
+  type: string;
+  data: T;
+}[] {
+  const messagesByRole = options?.role
+    ? messages.filter((msg) => msg.role === options?.role)
+    : messages;
+  const annotations = getAllAnnotations(messagesByRole);
+  const annotationsByType = options?.type
+    ? annotations.filter((a) => a.type === options.type)
+    : annotations;
+  return annotationsByType as { type: string; data: T }[];
+}

templates/components/llamaindex/typescript/documents/upload.ts

@@ -75,7 +75,7 @@ export function createCallbackManager(stream: StreamData) {
   callbackManager.on("retrieve-end", (data) => {
     const { nodes, query } = data.detail;
     appendSourceData(stream, nodes);
-    appendEventData(stream, `Retrieving context for query: '${query}'`);
+    appendEventData(stream, `Retrieving context for query: '${query.query}'`);
     appendEventData(
       stream,
       `Retrieved ${nodes.length} sources to use as context for the query`,

templates/components/llamaindex/typescript/streaming/annotations.ts

@@ -1,6 +1,5 @@
 import logging
 
-from app.api.routers.events import EventCallbackHandler
 from app.api.routers.models import (
     ChatData,
 )
@@ -23,13 +22,12 @@ async def chat(
         last_message_content = data.get_last_message_content()
         messages = data.get_history_messages(include_agent_messages=True)
 
-        event_handler = EventCallbackHandler()
         # The chat API supports passing private document filters and chat params
         # but agent workflow does not support them yet
         # ignore chat params and use all documents for now
         # TODO: generate filters based on doc_ids
-        # TODO: use chat params
-        engine = get_chat_engine(chat_history=messages)
+        params = data.data or {}
+        engine = get_chat_engine(chat_history=messages, params=params)
 
         event_handler = engine.run(input=last_message_content, streaming=True)
         return VercelStreamResponse(

templates/components/llamaindex/typescript/streaming/events.ts

@@ -1,6 +1,6 @@
+import asyncio
 import json
 import logging
-from abc import ABC
 from typing import AsyncGenerator, List
 
 from aiostream import stream
@@ -13,7 +13,7 @@
 logger = logging.getLogger("uvicorn")
 
 
-class VercelStreamResponse(StreamingResponse, ABC):
+class VercelStreamResponse(StreamingResponse):
     """
     Base class to convert the response from the chat engine to the streaming format expected by Vercel
     """
@@ -23,26 +23,34 @@ class VercelStreamResponse(StreamingResponse, ABC):
 
     def __init__(self, request: Request, chat_data: ChatData, *args, **kwargs):
         self.request = request
-
-        stream = self._create_stream(request, chat_data, *args, **kwargs)
-        content = self.content_generator(stream)
-
+        self.chat_data = chat_data
+        content = self.content_generator(*args, **kwargs)
         super().__init__(content=content)
 
-    async def content_generator(self, stream):
+    async def content_generator(self, event_handler, events):
+        logger.info("Starting content_generator")
+        stream = self._create_stream(
+            self.request, self.chat_data, event_handler, events
+        )
         is_stream_started = False
-
-        async with stream.stream() as streamer:
-            async for output in streamer:
-                if not is_stream_started:
-                    is_stream_started = True
-                    # Stream a blank message to start the stream
-                    yield self.convert_text("")
-
-                yield output
-
-                if await self.request.is_disconnected():
-                    break
+        try:
+            async with stream.stream() as streamer:
+                async for output in streamer:
+                    if not is_stream_started:
+                        is_stream_started = True
+                        # Stream a blank message to start the stream
+                        yield self.convert_text("")
+
+                    yield output
+        except asyncio.CancelledError:
+            logger.info("Stopping workflow")
+            await event_handler.cancel_run()
+        except Exception as e:
+            logger.error(
+                f"Unexpected error in content_generator: {str(e)}", exc_info=True
+            )
+        finally:
+            logger.info("The stream has been stopped!")
 
     def _create_stream(
         self,

templates/components/multiagent/python/app/api/routers/chat.py

@@ -18,11 +18,11 @@ def get_chat_engine(
     agent_type = os.getenv("EXAMPLE_TYPE", "").lower()
     match agent_type:
         case "choreography":
-            agent = create_choreography(chat_history)
+            agent = create_choreography(chat_history, **kwargs)
         case "orchestrator":
-            agent = create_orchestrator(chat_history)
+            agent = create_orchestrator(chat_history, **kwargs)
         case _:
-            agent = create_workflow(chat_history)
+            agent = create_workflow(chat_history, **kwargs)
 
     logger.info(f"Using agent pattern: {agent_type}")
 

templates/components/multiagent/python/app/api/routers/vercel_response.py

@@ -8,8 +8,8 @@
 from llama_index.core.chat_engine.types import ChatMessage
 
 
-def create_choreography(chat_history: Optional[List[ChatMessage]] = None):
-    researcher = create_researcher(chat_history)
+def create_choreography(chat_history: Optional[List[ChatMessage]] = None, **kwargs):
+    researcher = create_researcher(chat_history, **kwargs)
     publisher = create_publisher(chat_history)
     reviewer = FunctionCallingAgent(
         name="reviewer",
@@ -21,12 +21,14 @@ def create_choreography(chat_history: Optional[List[ChatMessage]] = None):
         name="writer",
         agents=[researcher, reviewer, publisher],
         description="expert in writing blog posts, needs researched information and images to write a blog post",
-        system_prompt=dedent("""
+        system_prompt=dedent(
+            """
             You are an expert in writing blog posts. You are given a task to write a blog post. Before starting to write the post, consult the researcher agent to get the information you need. Don't make up any information yourself.
             After creating a draft for the post, send it to the reviewer agent to receive feedback and make sure to incorporate the feedback from the reviewer.
             You can consult the reviewer and researcher a maximum of two times. Your output should contain only the blog post.
             Finally, always request the publisher to create a document (PDF, HTML) and publish the blog post.
-        """),
+        """
+        ),
         # TODO: add chat_history support to AgentCallingAgent
         # chat_history=chat_history,
     )

templates/components/multiagent/python/app/engine/engine.py

@@ -8,28 +8,32 @@
 from llama_index.core.chat_engine.types import ChatMessage
 
 
-def create_orchestrator(chat_history: Optional[List[ChatMessage]] = None):
-    researcher = create_researcher(chat_history)
+def create_orchestrator(chat_history: Optional[List[ChatMessage]] = None, **kwargs):
+    researcher = create_researcher(chat_history, **kwargs)
     writer = FunctionCallingAgent(
         name="writer",
         description="expert in writing blog posts, need information and images to write a post",
-        system_prompt=dedent("""
+        system_prompt=dedent(
+            """
             You are an expert in writing blog posts.
             You are given a task to write a blog post. Do not make up any information yourself.
             If you don't have the necessary information to write a blog post, reply "I need information about the topic to write the blog post".
             If you need to use images, reply "I need images about the topic to write the blog post". Do not use any dummy images made up by you.
             If you have all the information needed, write the blog post.
-        """),
+        """
+        ),
         chat_history=chat_history,
     )
     reviewer = FunctionCallingAgent(
         name="reviewer",
         description="expert in reviewing blog posts, needs a written blog post to review",
-        system_prompt=dedent("""
+        system_prompt=dedent(
+            """
             You are an expert in reviewing blog posts. You are given a task to review a blog post. Review the post and fix any issues found yourself. You must output a final blog post.
             A post must include at least one valid image. If not, reply "I need images about the topic to write the blog post". An image URL starting with "example" or "your website" is not valid.
             Especially check for logical inconsistencies and proofread the post for grammar and spelling errors.
-        """),
+        """
+        ),
         chat_history=chat_history,
     )
     publisher = create_publisher(chat_history)