openclaw - 💡(How to fix) Fix Context token count shows 0 for Ollama provider despite usage data being returned [1 comments, 2 participants]

Bug Description

When using the Ollama provider (ollama/phi3.5:latest), the session status shows 0/197k tokens (0%) even after active conversation. The context counter doesn't track token usage for Ollama models.

Root Cause

Ollama's OpenAI-compatible API (http://localhost:11434/v1/chat/completions) does return token usage data in the response:

"usage": { "prompt_tokens": 10, "completion_tokens": 10, "total_tokens": 20 }

However, OpenClaw's Ollama provider integration doesn't read/parse the usage field from Ollama's response, so the context display always shows 0.

Expected Behavior

Token usage should be tracked accurately for Ollama models. The model config correctly specifies contextWindow: 4096 for phi3.5, but the display falls back to the fallback model's 196K context.

Environment

• OpenClaw version: 2026.4.23
• Ollama version: running locally on http://localhost:11434
• Model: phi3.5:latest (3.8B, Q4_0, 4K context)
• GPU: NVIDIA 960M (4GB)

Model Config

"ollama": {
  "baseUrl": "http://localhost:11434/v1",
  "api": "openai-completions",
  "apiKey": "ollama",
  "models": [{
    "id": "phi3.5:latest",
    "name": "Phi-3.5 (Local GPU)",
    "contextWindow": 4096,
    "maxTokens": 2048,
    "input": ["text"],
    "cost": { "input": 0, "output": 0 }
  }]
}

Steps to Reproduce

1. Configure Ollama provider in openclaw.json
2. Switch to Ollama model via /model ollama/phi3.5:latest
3. Have an active conversation
4. Run /status — context shows 0 tokens despite active conversation

Suggested Fix

Ensure the Ollama provider parses the usage field from Ollama's API response. The data is already being returned; it's just not being read.