openclaw - ✅(Solved) Fix GitHub Copilot provider: missing Editor-Version header on inference requests causes HTTP 400 [2 pull requests, 1 comments, 2 participants]

mieurban · 2026-03-31T02:00:27Z

[openclaw] PR 62257: fix github-copilot : return IDE identity headers from prepareRuntimeAuth - Repository: openclaw/openclaw - Author: noahbclarkson - State:… # PR #62257: fix(github-copilot): return IDE identity headers from prepareRuntimeAuth - Repository: openclaw/openclaw - Author: noahbclarkson - State: open | merged: False - Link: https://github.com/openclaw/openclaw/pull/62257 ## Description (problem / solution / changelog) ## Summary Fixes #58056 — GitHub Copilot provider returns HTTP 400 `missing Editor-Version header for IDE auth` on all inference requests. The prior fix in #60755 added IDE identity headers (`Editor-Version`, `User-Agent`) to openclaw's transport layer (`copilot-dynamic-headers.ts`) and the extension's `wrapStreamFn`. However, **these code paths are bypassed during normal inference** because: 1. **Boundary-aware transport is never activated.** `pi-coding-agent` wraps `streamSimple` in a custom auth-injecting function during `createAgentSession`. The identity check in `stream-resolution.ts:116` (`currentStreamFn === streamSimple`) fails because it's a different function reference, so openclaw's transport layer — which has the correct headers — is never used. 2. **The actual inference path falls through to `@mariozechner/pi-ai`'s internal providers** (`anthropic.js`, `openai-responses.js`, `openai-completions.js`), which call pi-ai's own `buildCopilotDynamicHeaders` from `dist/providers/github-copilot-headers.js`. This is a **separate copy** that was never updated with the IDE headers — it only returns `X-Initiator`, `Openai-Intent`, and `Copilot-Vision-Request`. 3. **The extension's `wrapStreamFn` only covers `anthropic-messages` API** (`stream.ts:16`), leaving GPT-through-Copilot models (`openai-responses` API) with no IDE header injection at all. ## Fix Return the IDE headers from `prepareRuntimeAuth` via the existing `ProviderPreparedRuntimeAuth.request.headers` contract. This is a one-line addition: ```typescript request: { headers: buildCopilotIdeHeaders() }, ``` These headers flow through `applyPreparedRuntimeRequestOverrides` → `resolveProviderRequestConfig` → `model.headers`. Pi-ai's providers read `model.headers` early in their `mergeHeaders` chain, so the IDE headers reach the outgoing HTTP request **regardless of which stream function is active**. This fixes: - `anthropic-messages` API (Claude models through Copilot) - `openai-responses` API (GPT models through Copilot) - Both the main inference path and the compaction path Also exports `buildCopilotIdeHeaders` from `openclaw/plugin-sdk/provider-stream-shared` so the extension can use it without violating boundary rules. ## Remaining issues (not addressed here) - **`stream-resolution.ts:116` identity check**: The `=== streamSimple` check should be relaxed to also activate the boundary-aware transport for pi-coding-agent sessions. This would let openclaw's transport layer handle Copilot requests directly, bypassing pi-ai's providers entirely. - **`@mariozechner/pi-ai` stale headers**: Pi-ai's `buildCopilotDynamicHeaders` should eventually be updated to include IDE headers for defense in depth. - **Simple completion path**: Label generation and TTS use `streamSimple` without the `prepareRuntimeAuth` flow, so they remain affected. Lower severity since these are auxiliary tasks. ## Test plan - [ ] Verify `openclaw gateway run` with `github-copilot/claude-sonnet-4.6` no longer returns HTTP 400 - [ ] Verify GPT models through Copilot (`github-copilot/gpt-4o`) also work - [ ] Existing `extensions/github-copilot/stream.test.ts` continues to pass - [ ] `auth-controller.test.ts` tests for `applyPreparedRuntimeRequestOverrides` pass with `request.headers` 🤖 Generated with [Claude Code](https://claude.com/claude-code) ## Changed files - `extensions/github-copilot/index.ts` (modified, +2/-0) - `src/plugin-sdk/provider-stream-shared.ts` (modified, +1/-0) --- # PR #34: feat: add GitHub Copilot as a model provider - Repository: paoloanzn/free-code - Author: greatbody - State: open | merged: False - Link: https://github.com/paoloanzn/free-code/pull/34 ## Description (problem / solution / changelog) ## Summary - **Add GitHub Copilot as a new model provider**, integrating the Copilot API (`api.githubcopilot.com`) as an OpenAI-compatible backend. Supports chat completions, dynamic model listing via `/models` endpoint, and provider switching via `CLAUDE_CODE_USE_COPILOT=1` or `/provider copilot`. - **Register OpenAI env vars** (`OPENAI_API_KEY`, `OPENAI_BASE_URL`, etc.) in managed and safe env var sets, add `refreshModelStringsForCurrentProvider` utility, and include a test for the OpenAI-compatible fetch adapter. - **Rename package** from `claude-code-source-snapshot` to `free-code`. ## Usage ```bash CLAUDE_CODE_USE_COPILOT=1 COPILOT_TOKEN= ./cli ``` Or interactively: `/provider copilot` ## Files Changed - `src/utils/model/providers.ts` — Added `copilot` to `APIProvider` type - `src/utils/model/configs.ts` — Added `cop

openclaw2026-03-31 02:00:27

ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

GitHub issue URL

Helpful · Quick feedback

GitHub stats

openclaw/openclaw#58056•Fetched 2026-04-08 01:54:27

View on GitHub

Comments

Participants

Timeline

Reactions

Author

mieurban

Participants

erikrs92

mieurban

Timeline (top)

subscribed ×2commented ×1cross-referenced ×1referenced ×1

Error Message

Observe in logs: "error":"HTTP 400: bad request: missing Editor-Version header for IDE auth"

Root Cause

The Editor-Version header is correctly set in the auth/usage code paths:

usage-DLRV_xyV.js — "Editor-Version": "vscode/1.96.2" (usage endpoint)
auth-profiles-B5ypC5S-.js — "Editor-Version": "vscode/1.96.2" (token exchange)

However, the inference request path (the actual model API call via prepareRuntimeAuth in extensions/github-copilot/index.ts) only returns { apiKey, baseUrl, expiresAt } — there is no mechanism to inject custom headers into the outgoing model request.

The ProviderPreparedRuntimeAuth type only supports:

type ProviderPreparedRuntimeAuth = {
    apiKey: string;
    baseUrl?: string;
    expiresAt?: number;
};

There is no headers or extraHeaders field, so even though the GitHub Copilot API requires Editor-Version, Editor-Plugin-Version, and Copilot-Integration-Id headers, the plugin has no way to provide them through this hook.

Fix Action

Workaround

Adding headers via models.providers.github-copilot.headers in config partially works but requires a full provider entry with baseUrl and models[], which conflicts with the plugin's dynamic prepareRuntimeAuth base URL resolution. The config schema supports headers on provider entries, but the plugin-owned provider may not merge them.

PR fix notes

PR #62257: fix(github-copilot): return IDE identity headers from prepareRuntimeAuth

Repository: openclaw/openclaw
Author: noahbclarkson
State: open | merged: False
Link: https://github.com/openclaw/openclaw/pull/62257

Description (problem / solution / changelog)

Summary

Fixes #58056 — GitHub Copilot provider returns HTTP 400 missing Editor-Version header for IDE auth on all inference requests.

The prior fix in #60755 added IDE identity headers (Editor-Version, User-Agent) to openclaw's transport layer (copilot-dynamic-headers.ts) and the extension's wrapStreamFn. However, these code paths are bypassed during normal inference because:

Boundary-aware transport is never activated. pi-coding-agent wraps streamSimple in a custom auth-injecting function during createAgentSession. The identity check in stream-resolution.ts:116 (currentStreamFn === streamSimple) fails because it's a different function reference, so openclaw's transport layer — which has the correct headers — is never used.
The actual inference path falls through to @mariozechner/pi-ai's internal providers (anthropic.js, openai-responses.js, openai-completions.js), which call pi-ai's own buildCopilotDynamicHeaders from dist/providers/github-copilot-headers.js. This is a separate copy that was never updated with the IDE headers — it only returns X-Initiator, Openai-Intent, and Copilot-Vision-Request.
The extension's wrapStreamFn only covers anthropic-messages API (stream.ts:16), leaving GPT-through-Copilot models (openai-responses API) with no IDE header injection at all.

Fix

Return the IDE headers from prepareRuntimeAuth via the existing ProviderPreparedRuntimeAuth.request.headers contract. This is a one-line addition:

request: { headers: buildCopilotIdeHeaders() },

These headers flow through applyPreparedRuntimeRequestOverrides → resolveProviderRequestConfig → model.headers. Pi-ai's providers read model.headers early in their mergeHeaders chain, so the IDE headers reach the outgoing HTTP request regardless of which stream function is active.

This fixes:

anthropic-messages API (Claude models through Copilot)
openai-responses API (GPT models through Copilot)
Both the main inference path and the compaction path

Also exports buildCopilotIdeHeaders from openclaw/plugin-sdk/provider-stream-shared so the extension can use it without violating boundary rules.

Remaining issues (not addressed here)

stream-resolution.ts:116 identity check: The === streamSimple check should be relaxed to also activate the boundary-aware transport for pi-coding-agent sessions. This would let openclaw's transport layer handle Copilot requests directly, bypassing pi-ai's providers entirely.
@mariozechner/pi-ai stale headers: Pi-ai's buildCopilotDynamicHeaders should eventually be updated to include IDE headers for defense in depth.
Simple completion path: Label generation and TTS use streamSimple without the prepareRuntimeAuth flow, so they remain affected. Lower severity since these are auxiliary tasks.

Test plan

Verify openclaw gateway run with github-copilot/claude-sonnet-4.6 no longer returns HTTP 400
Verify GPT models through Copilot (github-copilot/gpt-4o) also work
Existing extensions/github-copilot/stream.test.ts continues to pass
auth-controller.test.ts tests for applyPreparedRuntimeRequestOverrides pass with request.headers

🤖 Generated with Claude Code

Changed files

extensions/github-copilot/index.ts (modified, +2/-0)
src/plugin-sdk/provider-stream-shared.ts (modified, +1/-0)

PR #34: feat: add GitHub Copilot as a model provider

Repository: paoloanzn/free-code
Author: greatbody
State: open | merged: False
Link: https://github.com/paoloanzn/free-code/pull/34

Description (problem / solution / changelog)

Summary

Add GitHub Copilot as a new model provider, integrating the Copilot API (api.githubcopilot.com) as an OpenAI-compatible backend. Supports chat completions, dynamic model listing via /models endpoint, and provider switching via CLAUDE_CODE_USE_COPILOT=1 or /provider copilot.
Register OpenAI env vars (OPENAI_API_KEY, OPENAI_BASE_URL, etc.) in managed and safe env var sets, add refreshModelStringsForCurrentProvider utility, and include a test for the OpenAI-compatible fetch adapter.
Rename package from claude-code-source-snapshot to free-code.

Usage

CLAUDE_CODE_USE_COPILOT=1 COPILOT_TOKEN=<github_oauth_token> ./cli

Or interactively: /provider copilot

Files Changed

src/utils/model/providers.ts — Added copilot to APIProvider type
src/utils/model/configs.ts — Added copilot field to all model configs
src/services/api/client.ts — Added Copilot client branch with Copilot-specific headers
src/services/api/bootstrap.ts — Added Copilot model refresh on startup
src/services/api/openaiCompatible.ts — New: OpenAI-compatible fetch adapter + refreshCopilotModelOptions
src/commands/provider/ — New: provider switching command with Copilot support
src/utils/model/model.ts — Added Copilot to default model selection
src/utils/model/modelOptions.ts — Added Copilot to model options
src/utils/status.tsx — Show Copilot base URL in /status
src/utils/managedEnvConstants.ts — Register OpenAI env vars
src/utils/model/modelStrings.ts — Add provider model refresh utility
package.json — Rename to free-code

Summary by CodeRabbit

New Features
- Added /provider command to switch between API providers: first-party Anthropic, Bedrock, Vertex AI, Foundry, OpenAI-compatible, and GitHub Copilot
- Extended model catalog with Copilot-compatible model identifiers
- Added support for OpenAI-compatible API endpoints
Package Changes
- Updated package name to "free-code"

Changed files

package.json (modified, +1/-1)
src/commands.ts (modified, +2/-0)
src/commands/provider/index.ts (added, +13/-0)
src/commands/provider/provider.ts (added, +180/-0)
src/services/api/bootstrap.ts (modified, +70/-56)
src/services/api/client.ts (modified, +170/-120)
src/services/api/openaiCompatible.test.ts (added, +161/-0)
src/services/api/openaiCompatible.ts (added, +657/-0)
src/utils/managedEnvConstants.ts (modified, +12/-0)
src/utils/model/configs.ts (modified, +108/-95)
src/utils/model/model.ts (modified, +255/-243)
src/utils/model/modelOptions.ts (modified, +239/-223)
src/utils/model/modelStrings.ts (modified, +17/-0)
src/utils/model/providers.ts (modified, +26/-18)
src/utils/status.tsx (modified, +248/-165)

Code Example

HTTP 400: bad request: missing Editor-Version header for IDE auth

---

type ProviderPreparedRuntimeAuth = {
    apiKey: string;
    baseUrl?: string;
    expiresAt?: number;
};

---

"github-copilot": {
  "claude-haiku-4.5": {
    headers: {
      "User-Agent": "GitHubCopilotChat/0.35.0",
      "Editor-Version": "vscode/1.107.0",
      "Editor-Plugin-Version": "copilot-chat/0.35.0",
      "Copilot-Integration-Id": "vscode-chat"
    }
  }
}

---

type ProviderPreparedRuntimeAuth = {
    apiKey: string;
    baseUrl?: string;
    expiresAt?: number;
    headers?: Record<string, string>;  // <-- new
};

---

prepareRuntimeAuth: async (ctx) => {
    const token = await resolveCopilotApiToken({
        githubToken: ctx.apiKey,
        env: ctx.env
    });
    return {
        apiKey: token.token,
        baseUrl: token.baseUrl,
        expiresAt: token.expiresAt,
        headers: {
            "Editor-Version": "vscode/1.107.0",
            "Editor-Plugin-Version": "copilot-chat/0.35.0",
            "Copilot-Integration-Id": "vscode-chat",
            "User-Agent": "GitHubCopilotChat/0.35.0"
        }
    };
},

RAW_BUFFERClick to expand / collapse

Bug Description

The GitHub Copilot provider plugin does not send the required Editor-Version header on model inference API requests, resulting in:

HTTP 400: bad request: missing Editor-Version header for IDE auth

Root Cause

The Editor-Version header is correctly set in the auth/usage code paths:

usage-DLRV_xyV.js — "Editor-Version": "vscode/1.96.2" (usage endpoint)
auth-profiles-B5ypC5S-.js — "Editor-Version": "vscode/1.96.2" (token exchange)

The ProviderPreparedRuntimeAuth type only supports:

type ProviderPreparedRuntimeAuth = {
    apiKey: string;
    baseUrl?: string;
    expiresAt?: number;
};

Evidence

The built-in model catalog (control-ui/assets/event-stream-*.js) does define the correct headers per-model:

"github-copilot": {
  "claude-haiku-4.5": {
    headers: {
      "User-Agent": "GitHubCopilotChat/0.35.0",
      "Editor-Version": "vscode/1.107.0",
      "Editor-Plugin-Version": "copilot-chat/0.35.0",
      "Copilot-Integration-Id": "vscode-chat"
    }
  }
}

But these catalog-defined headers are not applied to the actual inference fetch call, only to catalog metadata.

Reproduction

Install OpenClaw v2026.3.28 on a headless server (e.g., AWS Lightsail)
Run openclaw models auth login-github-copilot
Set model: openclaw models set github-copilot/claude-sonnet-4.6
Send a message via any channel (Telegram, web UI, etc.)
Observe in logs: "error":"HTTP 400: bad request: missing Editor-Version header for IDE auth"

Workaround

Suggested Fix

Option A (minimal): Extend ProviderPreparedRuntimeAuth to support an optional headers field:

type ProviderPreparedRuntimeAuth = {
    apiKey: string;
    baseUrl?: string;
    expiresAt?: number;
    headers?: Record<string, string>;  // <-- new
};

Then in extensions/github-copilot/index.ts, update prepareRuntimeAuth:

prepareRuntimeAuth: async (ctx) => {
    const token = await resolveCopilotApiToken({
        githubToken: ctx.apiKey,
        env: ctx.env
    });
    return {
        apiKey: token.token,
        baseUrl: token.baseUrl,
        expiresAt: token.expiresAt,
        headers: {
            "Editor-Version": "vscode/1.107.0",
            "Editor-Plugin-Version": "copilot-chat/0.35.0",
            "Copilot-Integration-Id": "vscode-chat",
            "User-Agent": "GitHubCopilotChat/0.35.0"
        }
    };
},

And update the inference fetch layer to merge headers from ProviderPreparedRuntimeAuth into the outgoing request.

Option B: Have the inference layer read per-model headers from the built-in catalog (which already defines them correctly) and apply them to the fetch call.

Environment

OpenClaw: v2026.3.28
OS: Ubuntu (AWS Lightsail)
Node: v22.22.2
Channel: Telegram + Web UI
Model: github-copilot/claude-sonnet-4.6 (also affects all other Copilot models)

extent analysis

Fix Plan

To resolve the issue, we will implement Option A (minimal): extend ProviderPreparedRuntimeAuth to support an optional headers field and update the prepareRuntimeAuth function to include the required headers.

Step 1: Update `ProviderPreparedRuntimeAuth` type

type ProviderPreparedRuntimeAuth = {
    apiKey: string;
    baseUrl?: string;
    expiresAt?: number;
    headers?: Record<string, string>;
};

Step 2: Update `prepareRuntimeAuth` function

prepareRuntimeAuth: async (ctx) => {
    const token = await resolveCopilotApiToken({
        githubToken: ctx.apiKey,
        env: ctx.env
    });
    return {
        apiKey: token.token,
        baseUrl: token.baseUrl,
        expiresAt: token.expiresAt,
        headers: {
            "Editor-Version": "vscode/1.107.0",
            "Editor-Plugin-Version": "copilot-chat/0.35.0",
            "Copilot-Integration-Id": "vscode-chat",
            "User-Agent": "GitHubCopilotChat/0.35.0"
        }
    };
},

Step 3: Update inference fetch layer to merge headers

// Assuming the fetch function is named 'fetchInference'
fetchInference: async (ctx, auth) => {
    const headers = auth.headers || {};
    // Merge headers from auth into the outgoing request
    const response = await fetch(ctx.baseUrl, {
        method: 'POST',
        headers: {
            ...headers,
            'Content-Type': 'application/json'
        },
        body: JSON.stringify(ctx.requestBody)
    });
    return response.json();
},

Verification

To verify the fix, follow these steps:

Update the code with the changes mentioned above.
Restart the OpenClaw server.
Run openclaw models auth login-github-copilot and set the model to github-copilot/claude-sonnet-4.6.
Send a message via any channel (Telegram, web UI, etc.).
Check the logs for any errors related to missing headers.

Extra Tips

Make sure to handle any potential errors that may occur during the fetch request.
Consider adding logging to track the headers being sent with each request.
If using a version control system, create a new branch for the changes and merge them into the main branch after verification.

Vote matrix · Quick signals

Works

Did the solution work? Tap to confirm.

Easy Fix

Was it a quick fix?

Time Saver

Did it save you time?

Blocking

Was it severely blocking?

Common Issue

Are others likely hitting this too?

Flaky / Intermittent

Is it intermittent?

Verified / Reproducible

Can you reproduce it reliably?

#api #ssr #installation #tensor shape #autograd error

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

openclaw - ✅(Solved) Fix GitHub Copilot provider: missing Editor-Version header on inference requests causes HTTP 400 [2 pull requests, 1 comments, 2 participants]

Recommended Tools

GitHub issue graph ai analysis

Error Message

Root Cause

Fix Action

Workaround

PR fix notes

PR #62257: fix(github-copilot): return IDE identity headers from prepareRuntimeAuth

Description (problem / solution / changelog)

Summary

Fix

Remaining issues (not addressed here)

Test plan

Changed files

PR #34: feat: add GitHub Copilot as a model provider

Description (problem / solution / changelog)

Summary

Usage

Files Changed

Summary by CodeRabbit

Changed files

Code Example

Bug Description

Root Cause

Evidence

Reproduction

Workaround

Suggested Fix

Environment

extent analysis

Fix Plan

Step 1: Update ProviderPreparedRuntimeAuth type

Step 2: Update prepareRuntimeAuth function

Step 3: Update inference fetch layer to merge headers

Verification

Extra Tips

Still need to ship something?

RELATED_DISCOVERY

TRENDING

Step 1: Update `ProviderPreparedRuntimeAuth` type

Step 2: Update `prepareRuntimeAuth` function