feat(cli-generator): send X-Fern platform-identification headers

[devin-ai-integration]

Description

Generated CLIs previously sent only a User-Agent. They now send the same Fern platform-identification headers the SDK generators use, plus the invoked command, on every HTTP request and WebSocket handshake:

X-Fern-SDK-Name: elevenlabs-cli
X-Fern-SDK-Version: 1.0.0
X-Fern-Language: Rust
X-Fern-CLI-Command: text-to-speech.convert

Requested in Slack ("I noticed that the CLI doesn't send the X-Fern-SDK-Name header, can we get that included?").

Changes Made

• sdk/src/http.rs: new HttpConfig::platform_headers() (X-Fern-Language: Rust, X-Fern-SDK-Name: <binary>-cli via the existing user_agent_product normalization, X-Fern-SDK-Version: CARGO_PKG_VERSION) and with_cli_command(Option<String>) which adds X-Fern-CLI-Command when the dispatch path knows the invoked command (blank/header-invalid values are dropped). build_client() installs them as reqwest default headers alongside the User-Agent.
• REST (openapi/binding.rs) and GraphQL (graphql/binding.rs) dispatch set the command from the invoked op path (_op_path.join(".")); AsyncAPI (asyncapi/binding.rs) sets it from channel.sdk_group_name ++ leaf_command_name, mirroring resolve_channel. Programmatic consumers without a known command send no X-Fern-CLI-Command.
• WebSocket handshakes don't go through reqwest, so build_handshake_request now takes the platform headers and inserts them before auth headers (same override semantics as User-Agent).
• Changelog entry at generators/cli/changes/unreleased/add-fern-platform-headers.yml.
• Regenerated seed/cli (second commit). Note: most of that diff is catch-up — the committed seed/cli snapshots were stale relative to main's generator (the Update Seed workflow has been failing on main), so the regeneration also pulls in unrelated drift (e.g. jsonl/raw/http format docs, Cargo.lock, types-crate output).
• Updated README.md generator (if applicable)

Testing

• Unit tests added/updated: platform_headers_identify_the_cli, platform_headers_include_the_invoked_command_when_set, with_cli_command_ignores_blank_or_invalid_values, a wire-level build_client_sends_platform_headers_on_the_wire (wiremock, asserts all four headers received), and handshake_request_sets_platform_headers for the WS path.
• Manual testing completed: cargo build --locked --all-features --tests and cargo test --locked --all-features in generators/cli/sdk (all 1908 tests pass); pnpm seed test --generator cli --skip-scripts — 151/151 fixtures pass.

Link to Devin session: https://app.devin.ai/sessions/5b7fd815c5074c19a4a85820accf7f56

---

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment, CI, and merge conflict monitoring

[devin-ai-integration]

Tested the platform headers end-to-end with real generated CLI binaries (no mocked runtime): built regenerated seed fixtures with cargo build --locked and ran them against local header-logging servers.

HTTP — allof-composition entities get-entity --base-url http://127.0.0.1:4010 (fixture seed/cli/allof): the mock server received x-fern-sdk-name: allof-composition-cli, x-fern-sdk-version: 0.0.0, x-fern-language: Rust, x-fern-cli-command: entities.get-entity. A second command (users list-users) correctly switched the command header to users.list-users.

WebSocket handshake (null-type fixture + AsyncApiBinding test harness)

Temporarily wired the sdk's agent.asyncapi.yaml into the null-type fixture's main.rs (reverted after testing) and ran null-type agent-messages --message '{"type":"pong"}' --base-url ws://127.0.0.1:4011 against a local WS echo server — the handshake carried all four headers, inserted next to the Upgrade headers:

Full demo recording in the Devin session: https://app.devin.ai/sessions/5b7fd815c5074c19a4a85820accf7f56

[devin-ai-integration]

Also tested against a customer's real ElevenLabs specs (their openapi.json — 268 paths — plus openapi-overrides.yml and generators.yml): generated with this branch's generator via docker-mode seed run, built with cargo build --locked, and ran the generated elevenlabs binary against a local header-logging mock server.

elevenlabs text-to-speech convert --voice-id testvoice --json '{"text":"hello"}' --base-url http://127.0.0.1:4010 → POST /v1/text-to-speech/testvoice received all four headers, and a 3-level command (conversational-ai agents list) correctly produced x-fern-cli-command: conversational-ai.agents.list:

• x-fern-sdk-name: elevenlabs-cli (from binaryName: elevenlabs)
• x-fern-sdk-version: 0.0.0 (generated crate's CARGO_PKG_VERSION)
• x-fern-language: Rust
• x-fern-cli-command: text-to-speech.convert / conversational-ai.agents.list (per invocation)

Note on WebSocket coverage for this binary

The customer's generators.yml references an AsyncAPI spec, but copySpecs.ts/renderMainRs currently only emits OpenApiBinding, so the generated CLI exposes no WS channel commands — WS handshake headers couldn't be exercised through this binary. They were verified separately with a hand-wired AsyncApiBinding harness (see earlier comment/evidence).

Full demo recording in the Devin session: https://app.devin.ai/sessions/5b7fd815c5074c19a4a85820accf7f56