Use DaoXE in OpenHands

OpenHands runs a full engineering loop — planning, file edits, terminal commands, PRs. Its docs state it connects to any LLM LiteLLM supports, and the OpenAI-compatible path is exactly DaoXE's shape (cheapest APIs guide): a base URL, a key, and a model ID with the right prefix.

Why route OpenHands through DaoXE#

• Agent-grade loops, your key. OpenHands' planner, editor and terminal tool calls all ride one provider — a DaoXE key covers hundreds of models from ~25 providers, so a strong model and a cheap one sit on the same balance.
• Two profiles, one bill. Named LLM configurations let a strong model handle planning while a budget model does routine edits — without a second account anywhere.
• Easy to audit. An agent's output is diffs and command runs — ideal ground truth for checking whether a cheap endpoint is serving you the real model (detect model swapping).

Set it up in OpenHands#

1. Open the advanced LLM panel. In the OpenHands UI go to Settings → LLM, click "see advanced settings" and enable the Advanced toggle — the Base URL field only exists there.
2. Prefix the model ID. Set Custom Model to openai/ + the exact ID from DaoXE's GET /v1/models — e.g. openai/YOUR_EXACT_MODEL_ID. OpenHands routes through LiteLLM, and openai/… is its provider prefix for any OpenAI-compatible server.
3. Point it at DaoXE. Base URL: https://daoxe.com/v1. API Key: your DaoXE key. Headless runs read the same trio via LLM_MODEL / LLM_BASE_URL / LLM_API_KEY.
4. Sanity-check the loop. Ask for a small real task — list files, patch a typo — and confirm the tool calls actually execute. If the agent "acts like a chatbot", the docs' own fix is a stronger model or a larger context window, not a settings hunt.

# Headless / CLI runs read the same trio
export LLM_MODEL="openai/YOUR_EXACT_MODEL_ID"   # the openai/ prefix is required
export LLM_BASE_URL="https://daoxe.com/v1"
export LLM_API_KEY="***"

Verify you actually get the model#

Prove the endpoint works before blaming the client — if this fails, no setting will fix it:

export DAOXE_API_KEY="your_api_key"

# List the exact model IDs your account can call
curl --fail-with-body --show-error --silent \
  https://daoxe.com/v1/models \
  -H "Authorization: Bearer ${DAOXE_API_KEY}"

Prove the endpoint outside the agent first, then diff a hard task against the official API at temperature 0:

curl --fail-with-body --show-error --silent \
  https://daoxe.com/v1/chat/completions \
  -H "Authorization: Bearer ${DAOXE_API_KEY}" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "YOUR_EXACT_MODEL_ID",
    "max_tokens": 64,
    "messages": [{"role": "user", "content": "Say hello in one sentence."}]
  }'

Frequently asked questions#