Stuck at Orchestrator

[fahdmirza]

Hi,
I am trying to review this tool for my channel here https://youtube.com/@fahdmirza , but having an issue getting it run. It gets installed fine, but when I give it a query, it just gets stuck at orchestration. Do I need to provide all of these API keys in every field? What is the minimum config to run this ?

Agentic Bench

AGENTIC_BENCH_MODEL_NAME=<text model name eg. "gpt-4o">
AGENTIC_BENCH_MODEL_BASE_URL=<text model base url eg. "https://api.openai.com/v1">
AGENTIC_BENCH_MODEL_API_KEY=
AGENTIC_BENCH_SUPPORTED_EXECUTORS='["Docker", "Local"]'
AGENTIC_BENCH_EXECUTOR="Local"

RAG Agent Configuration

LLAMA_API_KEY=<Obtain your API key by visiting LlamaIndex Cloud "https://llamaindex.cloud">
OPENAI_MODEL_NAME=<openai model name eg. "gpt-4o">
OPENAI_API_KEY=

Autoscrapper

AGENTIC_BROWSER_TEXT_MODEL=<text model name eg. "gpt-4o">
AGENTIC_BROWSER_TEXT_API_KEY=
AGENTIC_BROWSER_TEXT_BASE_URL=<text model base url eg. "https://api.openai.com/v1">

Screenshot Analysis Configuration

AGENTIC_BROWSER_SS_ENABLED=
AGENTIC_BROWSER_SS_MODEL=<screenshot model name eg. "gpt-4o">
AGENTIC_BROWSER_SS_API_KEY=
AGENTIC_BROWSER_SS_BASE_URL=<screenshot model base url eg. "https://api.openai.com/v1">

Google Search Configuration

GOOGLE_API_KEY=
GOOGLE_CX=

Browser Configuration

BROWSER_STORAGE_DIR=<path to browser storage directory eg. "./browser_storage">

Logging

LOGFIRE_TOKEN=

Frontend Websocket Configuration

VITE_WEBSOCKET_URL=<your websocket url eg. "ws://localhost:8081/ws">
VITE_MODEL_NAME=${AGENTIC_BENCH_MODEL_NAME} # either use existing ${AGENTIC_BENCH_MODEL_NAME} or give a text model name eg. "gpt-4o"

[aniketk13]

Hey @fahdmirza

The minimum config depends on what agent do you want to run
Considering you want to access all the Agents, all env variables are required.

May I know which ones did you fill and which were not filled so I can get a better understanding?
Also if you could share the desired query, I can guide you the minimum config based on the requried agent.

[fahdmirza]

Hi,
I have filled in OPENAI API KEY, LLAMA CLOUD . I am just trying to do a simple web search. I believe in its current form it is too cumbersome a configuration. There are too many API keys to be honest. Can you please give me a simple minimal configuration which would work with a query like, "what is current stock price of NVIDIA" or "Who is Fahd Mirza from Australia?"

regards,
Fahd

[malv-c]

fahd is right.

• no local llm support.

[prakharsr]

@fahdmirza These are the minimum keys needed for running a query like "what is the current stock price of NVIDIA". Please refer to the .env.sample for explanation and default values of these keys.

• Agentic Bench, needed. Can give any openai compatible endpoint. Supports local llm also
  AGENTIC_BENCH_MODEL_NAME, AGENTIC_BENCH_MODEL_BASE_URL, AGENTIC_BENCH_MODEL_API_KEY, AGENTIC_BENCH_SUPPORTED_EXECUTORS, AGENTIC_BENCH_EXECUTOR

• Agentic Browser, needed. Can give any openai compatible endpoint. Supports local llm also
  AGENTIC_BROWSER_TEXT_MODEL, AGENTIC_BROWSER_TEXT_API_KEY, AGENTIC_BROWSER_TEXT_BASE_URL

• Screenshot Analysis Configuration, keep it as false and skip all other keys in this subsection
  AGENTIC_BROWSER_SS_ENABLED=false

• Google Search Configuration, needed
  GOOGLE_API_KEY, GOOGLE_CX

• Browser Configuration, static config, needed. Can keep as default
  BROWSER_STORAGE_DIR

• Frontend Websocket Configuration. Static config, needed. Can keep as default
  VITE_WEBSOCKET_URL, VITE_MODEL_NAME

@malv-c Local llms with a compatible openai endpoint are also supported.