[dawn-pixellate]

Adds Fikra to the OpenAI API compatible provider list in docs/customize/model-providers/top-level/openai.mdx and adds a concrete ### Fikra API example below the generic apiBase section, matching the existing YAML/JSON Tabs format.

The example uses provider: openai with apiBase: https://fikraapi.co.ke/v1 and model deepseek-v4.1-flash.

Facts kept to Fikra's published copy, DeepSeek first:

• DeepSeek V4.1 Flash: $0.045 per 1M input / $0.09 per 1M output, 1M context
• Also served: GLM 5.3, GLM 5.3 Flash, GLM 5.2, Kimi K3, StepFun Step 5, Qwen 3.8 27B (cached input $0.10/M), Xiaomi MiMo V2.6 Pro, NVIDIA Nemotron 3.5 Super VL, plus Fikra house lanes
• Full catalog at https://fikraapi.co.ke/v1/models; keys and billing (per-token USD, $1 minimum top-up, M-Pesa and card) at https://fikraapi.co.ke

[github-actions]

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

[dawn-pixellate]

I have read the CLA Document and I hereby sign the CLA

[dawn-pixellate]

This PR changes one docs page. The failing jetbrains-tests job dies inside AnimMouse/setup-ffmpeg@v1 while downloading FFmpeg 7.1 (xz: (stdin): File format not recognized), before any Continue code runs. It failed the same way on both the 10-19 and 10-25 runs, so it looks like the action's download, not a change here.

[dawn-pixellate]

Updating with the full check status on b711ef6, all three reds are CI infrastructure, not this PR:

• jetbrains-tests: AnimMouse/setup-ffmpeg@v1 fails while downloading FFmpeg 7.1 (xz: (stdin): File format not recognized), before any repo code runs. Failed identically on the 10-19 and 10-25 runs.
• e2e/_output/tests/Edit.test.js (e2e:ci:run-yaml): npm @vscode/ripgrep postinstall got 403 from api.github.com (rate limit) while fetching ripgrep-prebuilt v13.0.0-10. The exact same test passed in the e2e:ci:run job of the same run.
• security/snyk: third-party app status stuck in error state.
• require-all-checks-to-pass: just the gate aggregating the above.

The PR itself is a 39-line addition to docs/customize/model-providers/top-level/openai.mdx, and this head passed the e2e suite where the runner had rate-limit headroom.