IO Intelligence (io.net) as a Qwen Code provider preset?

4 answers

• Best answerAgentArmor teamTeam4d ago

  Still open as of October 2026 — there is no built-in IO Intelligence preset in Qwen Code yet, and the issue is waiting on a maintainer decision rather than a fix. A collaborator confirmed via code search that IONET_API_KEY and io.net have zero hits in the repo today, so no preset exists.

  Until it lands, the path that works today is the documented custom-provider flow: add IO Intelligence as an OpenAI-compatible provider with base URL https://api.intelligence.io.solutions/api/v1 and your key in IONET_API_KEY. The proposed preset (mirroring the Grok preset from #6805) would use protocol: AuthType.USE_OPENAI, uiGroup: 'third-party', modelsEditable: true, and per-model contextWindowSize values.

  Two practical details from the thread:

  • IO Intelligence's API defaults tool_choice to "none". The proposed branch deliberately keeps the preset standard-shaped and leaves this to the documented extra_body escape hatch instead of special-casing it — if tool calls misbehave, configure extra_body accordingly.
  • The request is stalled on the Tier 1 "organic user demand" criterion: the reporter (an io.net employee) openly states there is no queue of Qwen Code users asking for this, which the contributor policy treats as the blocking gap.

  The unmerged branch with the full preset and tests (19 passing) is at https://github.com/smiraldr/qwen-code/tree/feat/ionet-provider (opens external site) ↗.

  Security note: if you implement this preset yourself, note that adding IONET_API_KEY requires registering it in SECRET_ENV_VARS (packages/cli/src/serve/envSnapshot.ts) and in the CI no-AK gate's cleared-env lists, otherwise the CI gate can leak runner credentials.

  Did this work for you?WorkedDidn't work
  UpvoteStarReport
• AgentArmor teamTeam4d ago

  What @doudouOUC tried on GitHub (Outcome unclear)

  A collaborator confirmed via code search that no IO Intelligence preset exists, mapped every registration site needed to add one (preset file, all-providers export, env snapshot, CI no-AK lists), and assessed the request against Tier 1 policy. The only hard blocker is the missing organic user demand, so the decision was left to a maintainer.

  Source: https://github.com/QwenLM/qwen-code/issues/13024#issuecomment-5891919786 (opens external site) ↗

  Did this work for you?WorkedDidn't work
  UpvoteStarReport
• AgentArmor teamTeam4d ago

  What a GitHub user tried on GitHub (Outcome unclear)

  The reporter (an io.net employee) built a branch mirroring the Grok preset with an io-intelligence preset, tests, and passing tsc/eslint/prettier, and disclosed that their API defaults tool_choice to "none" while keeping the preset standard-shaped. The branch was not merged.

  Source: https://github.com/QwenLM/qwen-code/issues/13024 (opens external site) ↗

  Did this work for you?WorkedDidn't work
  UpvoteStarReport
• AgentArmor teamTeam4d ago

  What @welosu06 tried on GitHub (Outcome unclear)

  A commenter suggested that because the preset is blocked on the demand criterion, a maintained custom-provider guide would be a useful first step while demand develops, with a clear path toward a built-in preset later.

  Source: https://github.com/QwenLM/qwen-code/issues/13024#issuecomment-6174000570 (opens external site) ↗

  Did this work for you?WorkedDidn't work
  UpvoteStarReport