Copy page

Connect your coding agent

Aider

Aider is a terminal pair programmer that edits files in your git repo and commits as it goes. It connects to any OpenAI-compatible endpoint.

Setup guide: this config comes from Aider's official documentation and has not been run end to end against the gateway. Corrections to support@hyphen-solution.com.

The easy mistake

Aider routes through LiteLLM, which needs to be told what kind of endpoint it is talking to. That is what the openai/ prefix on the model name does.

Code
aider --model openai/minimax-m3     # correct
aider --model minimax-m3            # fails, LiteLLM cannot route it

The prefix is stripped before the request, so Hyphen still receives minimax-m3. Also note the env var is OPENAI_API_BASE, not OPENAI_BASE_URL. Aider does not read the latter.

1. Get a key

app.hyphen-solution.com → sign in → pick a plan or credit pack → Dashboard → API Keys. Copy the sk-... key.

2. Configure

Code
export OPENAI_API_BASE="https://api.hyphen-solution.com/v1"
export OPENAI_API_KEY="sk-YOUR_KEY"

aider --model openai/minimax-m3

On Windows use setx and restart the shell:

Code
setx OPENAI_API_BASE https://api.hyphen-solution.com/v1
setx OPENAI_API_KEY sk-YOUR_KEY

As flags

Code
aider \
  --openai-api-base https://api.hyphen-solution.com/v1 \
  --openai-api-key sk-YOUR_KEY \
  --model openai/minimax-m3

As a config file

.aider.conf.yml in your home directory or git root:

Code
openai-api-base: https://api.hyphen-solution.com/v1
openai-api-key: sk-YOUR_KEY
model: openai/minimax-m3
weak-model: openai/minimax-m2.5

weak-model is what Aider uses for commit messages and summarising chat history. Point it at minimax-m2.5 so those small jobs finish fast.

As a .env file

Aider reads .env with AIDER_-prefixed names:

Code
AIDER_OPENAI_API_BASE=https://api.hyphen-solution.com/v1
AIDER_OPENAI_API_KEY=sk-YOUR_KEY
AIDER_MODEL=openai/minimax-m3
AIDER_WEAK_MODEL=openai/minimax-m2.5

Plain OPENAI_API_BASE and OPENAI_API_KEY also work in .env.

3. Verify

Code
cd your-repo
aider --model openai/minimax-m3

At the prompt, ask it to reply with a single word. Aider prints the model it is using on startup, so check that line says openai/minimax-m3.

Telling Aider the context window and cost

Hyphen models are not in LiteLLM's model database, so Aider does not know their context window or price and will say so. Fix it with a metadata file. Create .aider.model.metadata.json in your home directory or git root:

Code
{
  "openai/minimax-m3": {
    "max_tokens": 32000,
    "max_input_tokens": 200000,
    "max_output_tokens": 32000,
    "input_cost_per_token": 0.0000003,
    "output_cost_per_token": 0.0000012,
    "litellm_provider": "openai",
    "mode": "chat"
  },
  "openai/minimax-m2.5": {
    "max_tokens": 32000,
    "max_input_tokens": 200000,
    "max_output_tokens": 32000,
    "input_cost_per_token": 0.0000003,
    "output_cost_per_token": 0.0000012,
    "litellm_provider": "openai",
    "mode": "chat"
  },
  "openai/minimax-text-01": {
    "max_tokens": 32000,
    "max_input_tokens": 4000000,
    "max_output_tokens": 32000,
    "input_cost_per_token": 0.0000002,
    "output_cost_per_token": 0.0000011,
    "litellm_provider": "openai",
    "mode": "chat"
  }
}

Key each entry by the fully qualified name including the openai/ prefix. Those costs are the real rate card, so Aider's running total will match your console. Point at a file elsewhere with --model-metadata-file path/to/file.json.

Aider only reports token limits, it does not enforce them, so this file is optional. Without it you lose the cost display and get a warning on startup.

Which model

• openai/minimax-m3 as the main model. Aider does whole-file and diff edits that reward the flagship.
• openai/minimax-m2.5 as the weak-model for commit messages.
• openai/minimax-text-01 if you use /add on a very large set of files.

Troubleshooting

• "LLM Provider NOT provided" or a routing error. You left off the openai/ prefix.
• 404. OPENAI_API_BASE is missing the /v1 suffix, or you set OPENAI_BASE_URL instead, which Aider ignores.
• 401. Wrong key.
• Warning about unknown context window. Add the metadata file above.
• Empty or truncated edits. The M-series reason before answering. See Choosing a model.
• 429. Monthly budget spent. See Rate limits & caps.
• Flags changed? Check Aider's OpenAI-compatible docs and the options reference.

Related

• Choosing a model: which model for which job.
• Other OpenAI-compatible clients: opencode, Cursor, Cline, Continue.