Skip to content

Common Issues

Fixes for the problems you are most likely to hit when installing and running Buddy AI. See also Debugging and the FAQ.

Installation & extras

Buddy keeps optional dependencies out of the base install. If a provider, tool, or feature reports a missing module, install the matching extra.

Symptom Cause Fix
ImportError: No module named 'anthropic' Anthropic provider not installed pip install "buddy-ai[anthropic]"
ImportError: ... 'google.generativeai' Google provider missing pip install "buddy-ai[google]"
ImportError: No module named 'boto3' AWS Bedrock missing pip install "buddy-ai[aws]"
ImportError: No module named 'torch' Training extras missing pip install "buddy-ai[training]"
ImportError: No module named 'PIL' / cv2 Multimodal extras missing pip install "buddy-ai[multimodal]"
ImportError: The 'chromadb' package is not installed Vector DB missing pip install chromadb

Install everything

To pull in all common providers and features at once: 

pip install "buddy-ai[all]"
Available extras include openai, anthropic, google, cohere, aws, azure, irag, multimodal, training, tools, and all.

Missing API keys

Most providers read their key from an environment variable. A missing key usually surfaces as an authentication error from the provider SDK.

export OPENAI_API_KEY=sk-...
export ANTHROPIC_API_KEY=sk-ant-...
export GOOGLE_API_KEY=AIza...
export TAVILY_API_KEY=tvly-...   # for TavilyTools

Verify a key is visible to your process:

echo $OPENAI_API_KEY

Local models need no key

Providers like Ollama (from buddy.models.ollama import Ollama) run locally and require no API key.

Wrong model id

Each provider validates the id you pass. A typo or an id your account cannot access raises a provider error. Double-check the id against your provider's catalog:

from buddy.models.openai import OpenAIChat

agent = Agent(model=OpenAIChat(id="gpt-4o-mini"))  # not "gpt4o-mini"

Agent returns an empty response

  1. Confirm the API key is set (see above).
  2. Turn on debug logging to see the raw request/response: 
    agent = Agent(model=OpenAIChat(id="gpt-4o-mini"), debug_mode=True)
  3. Inspect the returned object directly: print(agent.run("hi").to_dict()).

Rate-limit / transient errors

Add retries at the model level, or per run:

from buddy.models.openai import OpenAIChat

agent = Agent(model=OpenAIChat(id="gpt-4o-mini", max_retries=3))
# or bound retries on a single call:
agent.run("Summarize this.", retries=2)

The agent also supports retries, delay_between_retries, and exponential_backoff constructor parameters.

Memory not persisting between runs

Pass a stable session_id (and user_id) across calls, and configure a backend so state survives process restarts:

agent.run("Hello", session_id="my-session", user_id="user-123")
agent.run("Remember me?", session_id="my-session", user_id="user-123")

For durable storage use storage=SqliteStorage(...) for sessions and memory=Memory(db=SqliteMemoryDb(...)) for user memories — see the memory example.

Optional feature not available

If a feature like planning or PULSE seems missing, its extra may not be installed. Check at runtime:

from buddy import get_available_features, check_feature

print(get_available_features())   # e.g. ["reasoning", "pulse", "core"]
print(check_feature("planning"))  # False if its deps are missing