Adding web search infra

docker-compose.dev.yml

@@ -40,6 +40,7 @@ services:
       - FLASK_ENV=development
       - PYTHONUNBUFFERED=1
       - NODE_ENV=development
+      - TAVILY_KEY=${TAVILIY_KEY}
     depends_on:
       postgres:
         condition: service_healthy

services/raggr/blueprints/conversation/__init__.py

@@ -12,7 +12,6 @@ from .agents import main_agent
 from .logic import (
     add_message_to_conversation,
     get_conversation_by_id,
-    get_conversation_transcript,
     rename_conversation,
 )
 from .models import (
@@ -43,22 +42,29 @@ async def query():
         user=user,
     )
 
-    transcript = await get_conversation_transcript(user=user, conversation=conversation)
+    # Build conversation history from recent messages (last 10 for context)
+    recent_messages = (
+        conversation.messages[-10:]
+        if len(conversation.messages) > 10
+        else conversation.messages
+    )
 
-    transcript_prompt = f"Here is the message transcript thus far {transcript}."
+    messages_payload = [
-    prompt = f"""Answer the user in as if you were a cat named Simba. Don't act too catlike. Be assertive.
+        {
-    {transcript_prompt if len(transcript) > 0 else ""}
+            "role": "system",
-    Respond to this prompt: {query}"""
+            "content": "You are a helpful cat assistant named Simba that understands veterinary terms. When there are questions to you specifically, they are referring to Simba the cat. Answer the user in as if you were a cat named Simba. Don't act too catlike. Be assertive.\n\nIMPORTANT: When users ask factual questions about Simba's health, medical history, veterinary visits, medications, weight, or any information that would be in documents, you MUST use the simba_search tool to retrieve accurate information before answering. Do not rely on general knowledge - always search the documents for factual questions.",
+        }
+    ]
 
-    payload = {
+    # Add recent conversation history
-        "messages": [
+    for msg in recent_messages[:-1]:  # Exclude the message we just added
-            {
+        role = "user" if msg.speaker == "user" else "assistant"
-                "role": "system",
+        messages_payload.append({"role": role, "content": msg.text})
-                "content": "You are a helpful cat assistant named Simba that understands veterinary terms. When there are questions to you specifically, they are referring to Simba the cat. Answer the user in as if you were a cat named Simba. Don't act too catlike. Be assertive.\n\nIMPORTANT: When users ask factual questions about Simba's health, medical history, veterinary visits, medications, weight, or any information that would be in documents, you MUST use the simba_search tool to retrieve accurate information before answering. Do not rely on general knowledge - always search the documents for factual questions.",
+
-            },
+    # Add current query
-            {"role": "user", "content": prompt},
+    messages_payload.append({"role": "user", "content": query})
-        ]
+
-    }
+    payload = {"messages": messages_payload}
 
     response = await main_agent.ainvoke(payload)
     message = response.get("messages", [])[-1].content

services/raggr/blueprints/conversation/agents.py

@@ -1,10 +1,52 @@
+import os
+from typing import cast
+
 from langchain.agents import create_agent
+from langchain.chat_models import BaseChatModel
 from langchain.tools import tool
+from langchain_ollama import ChatOllama
 from langchain_openai import ChatOpenAI
+from tavily import AsyncTavilyClient
 
 from blueprints.rag.logic import query_vector_store
 
 openai_gpt_5_mini = ChatOpenAI(model="gpt-5-mini")
+ollama_deepseek = ChatOllama(model="llama3.1:8b", base_url=os.getenv("OLLAMA_URL"))
+model_with_fallback = cast(
+    BaseChatModel, ollama_deepseek.with_fallbacks([openai_gpt_5_mini])
+)
+client = AsyncTavilyClient(os.getenv("TAVILY_KEY"), "")
+
+
+@tool
+async def web_search(query: str) -> str:
+    """Search the web for current information using Tavily.
+
+    Use this tool when you need to:
+    - Find current information not in the knowledge base
+    - Look up recent events, news, or updates
+    - Verify facts or get additional context
+    - Search for information outside of Simba's documents
+
+    Args:
+        query: The search query to look up on the web
+
+    Returns:
+        Search results from the web with titles, content, and source URLs
+    """
+    response = await client.search(query=query, search_depth="basic")
+    results = response.get("results", [])
+
+    if not results:
+        return "No results found for the query."
+
+    formatted = "\n\n".join(
+        [
+            f"**{result['title']}**\n{result['content']}\nSource: {result['url']}"
+            for result in results[:5]
+        ]
+    )
+    return formatted
 
 
 @tool(response_format="content_and_artifact")
@@ -33,4 +75,4 @@ async def simba_search(query: str):
     return serialized, docs
 
 
-main_agent = create_agent(model=openai_gpt_5_mini, tools=[simba_search])
+main_agent = create_agent(model=model_with_fallback, tools=[simba_search, web_search])

services/raggr/blueprints/rag/logic.py

@@ -8,7 +8,7 @@ from langchain_text_splitters import RecursiveCharacterTextSplitter
 
 from .fetchers import PaperlessNGXService
 
-embeddings = OpenAIEmbeddings(model="text-embedding-3-large")
+embeddings = OpenAIEmbeddings(model="text-embedding-3-small")
 
 vector_store = Chroma(
     collection_name="simba_docs",
@@ -61,7 +61,7 @@ async def index_documents():
 
 
 async def query_vector_store(query: str):
-    retrieved_docs = vector_store.similarity_search(query, k=2)
+    retrieved_docs = await vector_store.asimilarity_search(query, k=2)
     serialized = "\n\n".join(
         (f"Source: {doc.metadata}\nContent: {doc.page_content}")
         for doc in retrieved_docs

services/raggr/pyproject.toml

@@ -34,6 +34,8 @@ dependencies = [
     "langchain-chroma>=1.0.0",
     "langchain-community>=0.4.1",
     "jq>=1.10.0",
+    "langchain-ollama>=1.0.1",
+    "tavily-python>=0.7.17",
 ]
 
 [tool.aerich]

services/raggr/uv.lock

@@ -1281,6 +1281,19 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/83/bd/9df897cbc98290bf71140104ee5b9777cf5291afb80333aa7da5a497339b/langchain_core-1.2.5-py3-none-any.whl", hash = "sha256:3255944ef4e21b2551facb319bfc426057a40247c0a05de5bd6f2fc021fbfa34", size = 484851, upload-time = "2025-12-22T23:45:30.525Z" },
 ]
 
+[[package]]
+name = "langchain-ollama"
+version = "1.0.1"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "langchain-core" },
+    { name = "ollama" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/73/51/72cd04d74278f3575f921084f34280e2f837211dc008c9671c268c578afe/langchain_ollama-1.0.1.tar.gz", hash = "sha256:e37880c2f41cdb0895e863b1cfd0c2c840a117868b3f32e44fef42569e367443", size = 153850, upload-time = "2025-12-12T21:48:28.68Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/e3/46/f2907da16dc5a5a6c679f83b7de21176178afad8d2ca635a581429580ef6/langchain_ollama-1.0.1-py3-none-any.whl", hash = "sha256:37eb939a4718a0255fe31e19fbb0def044746c717b01b97d397606ebc3e9b440", size = 29207, upload-time = "2025-12-12T21:48:27.832Z" },
+]
+
 [[package]]
 name = "langchain-openai"
 version = "1.1.6"
@@ -2521,6 +2534,7 @@ dependencies = [
     { name = "langchain" },
     { name = "langchain-chroma" },
     { name = "langchain-community" },
+    { name = "langchain-ollama" },
     { name = "langchain-openai" },
     { name = "ollama" },
     { name = "openai" },
@@ -2533,6 +2547,7 @@ dependencies = [
     { name = "python-dotenv" },
     { name = "quart" },
     { name = "quart-jwt-extended" },
+    { name = "tavily-python" },
     { name = "tomlkit" },
     { name = "tortoise-orm" },
     { name = "tortoise-orm-stubs" },
@@ -2554,6 +2569,7 @@ requires-dist = [
     { name = "langchain", specifier = ">=1.2.0" },
     { name = "langchain-chroma", specifier = ">=1.0.0" },
     { name = "langchain-community", specifier = ">=0.4.1" },
+    { name = "langchain-ollama", specifier = ">=1.0.1" },
     { name = "langchain-openai", specifier = ">=1.1.6" },
     { name = "ollama", specifier = ">=0.6.0" },
     { name = "openai", specifier = ">=2.0.1" },
@@ -2566,6 +2582,7 @@ requires-dist = [
     { name = "python-dotenv", specifier = ">=1.0.0" },
     { name = "quart", specifier = ">=0.20.0" },
     { name = "quart-jwt-extended", specifier = ">=0.1.0" },
+    { name = "tavily-python", specifier = ">=0.7.17" },
     { name = "tomlkit", specifier = ">=0.13.3" },
     { name = "tortoise-orm", specifier = ">=0.25.1" },
     { name = "tortoise-orm-stubs", specifier = ">=1.0.2" },
@@ -2847,6 +2864,20 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/a2/09/77d55d46fd61b4a135c444fc97158ef34a095e5681d0a6c10b75bf356191/sympy-1.14.0-py3-none-any.whl", hash = "sha256:e091cc3e99d2141a0ba2847328f5479b05d94a6635cb96148ccb3f34671bd8f5", size = 6299353, upload-time = "2025-04-27T18:04:59.103Z" },
 ]
 
+[[package]]
+name = "tavily-python"
+version = "0.7.17"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "httpx" },
+    { name = "requests" },
+    { name = "tiktoken" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/5f/eb/d7371ee68119380ab6561c6998eacf3031327ba89c6081d36128ab4a2184/tavily_python-0.7.17.tar.gz", hash = "sha256:437ba064639dfdce1acdbc37cbb73246abe500ab735e988a4b8698a8d5fb7df7", size = 21321, upload-time = "2025-12-17T17:08:39.3Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/c7/ce/88565f0c9f7654bc90e19f1e76b3bffee7ff9c1741a2124ec2f2900fb080/tavily_python-0.7.17-py3-none-any.whl", hash = "sha256:a2725b9cba71e404e73d19ff277df916283c10100137c336e07f8e1bd7789fcf", size = 18214, upload-time = "2025-12-17T17:08:38.442Z" },
+]
+
 [[package]]
 name = "tenacity"
 version = "9.1.2"