Add --anthropic-path-prefix flag and auto-probe for Envoy AI Gateway

[yrobla]

Summary

Some gateways (e.g. Envoy AI Gateway) serve native-Anthropic traffic at /anthropic/v1/messages rather than /v1/messages, causing Claude Code to receive 404s and report "model not found" errors.

• Add --anthropic-path-prefix flag to thv llm setup (default empty). When set, the prefix is appended to the gateway URL when writing ANTHROPIC_BASE_URL in direct-mode tool configs (Claude Code, Codex).
• Auto-probe: if the flag is omitted, issue a HEAD request to /anthropic/v1/messages; a 401 response (auth-reachable path) triggers automatic prefix selection. 404 or any error means no prefix.
• Proxy-mode tools (Cursor, VS Code) are unaffected — they route through thv llm proxy and don't use ANTHROPIC_BASE_URL.

Fixes #5158

Type of change

• Bug fix
• New feature
• Refactoring (no behavior change)
• Dependency update
• Documentation
• Other (describe):

Test plan

• Unit tests (task test)
• E2E tests (task test-e2e)
• Linting (task lint-fix)
• Manual testing (describe below)

API Compatibility

• This PR does not break the v1beta1 API, OR the api-break-allowed label is applied and the migration guidance is described above.

Changes

File	Change

Does this introduce a user-facing change?

Implementation plan

Approved implementation plan

Special notes for reviewers

[Copilot]

Pull request overview

Adds support in thv llm setup for gateways (notably Envoy AI Gateway) that serve Anthropic-native traffic under an /anthropic path prefix, preventing 404 “model not found” errors for direct-mode tools that rely on ANTHROPIC_BASE_URL.

Changes:

• Introduces --anthropic-path-prefix and propagates an effective AnthropicBaseURL into direct-mode tool config patching.
• Adds an auto-probe that HEADs <gateway>/anthropic/v1/messages and selects /anthropic when a 401 indicates the route exists.
• Updates Claude Code integration to write ANTHROPIC_BASE_URL from AnthropicBaseURL (with fallback to GatewayURL) and adds unit tests for prefix behavior + probing.

Reviewed changes

Copilot reviewed 7 out of 7 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
pkg/llmgateway/config.go	Extends ApplyConfig to carry an AnthropicBaseURL alongside the gateway URL.
pkg/llm/setup.go	Resolves/passes the Anthropic prefix, applies it only for direct-mode tools, and implements the auto-probe.
pkg/llm/setup_test.go	Adds unit tests for direct/proxy prefix behavior and probe outcomes.
pkg/llm/manage.go	Extends SetOptions with AnthropicPathPrefix.
pkg/client/llm_gateway.go	Adds AnthropicBaseURL value resolution with fallback to GatewayURL.
pkg/client/config.go	Switches Claude Code’s ANTHROPIC_BASE_URL mapping to AnthropicBaseURL.
cmd/thv/app/llm.go	Adds the --anthropic-path-prefix CLI flag to thv llm setup.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[codecov]

Codecov Report

❌ Patch coverage is 63.33333% with 22 lines in your changes missing coverage. Please review.
✅ Project coverage is 67.57%. Comparing base (9057f7a) to head (965c18c).

Files with missing lines	Patch %	Lines
pkg/llm/setup.go	64.28%	14 Missing and 6 partials ⚠️
pkg/client/llm_gateway.go	50.00%	1 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #5174      +/-   ##
==========================================
- Coverage   67.57%   67.57%   -0.01%     
==========================================
  Files         602      602              
  Lines       61119    61167      +48     
==========================================
+ Hits        41303    41334      +31     
- Misses      16693    16704      +11     
- Partials     3123     3129       +6     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Copilot]

Pull request overview

Copilot reviewed 7 out of 7 changed files in this pull request and generated 3 comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 7 out of 7 changed files in this pull request and generated 1 comment.

Comments suppressed due to low confidence (1)

cmd/thv/app/llm.go:233

• PR description mentions Codex being affected (“direct-mode tool configs (Claude Code, Codex)”), but Codex doesn’t appear to have LLM gateway support in supportedClientIntegrations (it has no LLMGatewayMode / LLMGatewayKeys). If Codex is intentionally out of scope, please update the PR description to avoid confusing users; if it should be in scope, it likely needs an LLM gateway config entry similar to other direct-mode tools.

	cmd := &cobra.Command{
		Use:   "setup",
		Short: "Configure detected AI tools to use the LLM gateway",
		Long: `Detect installed AI coding tools (Claude Code, Gemini CLI, Cursor, VS Code,
Xcode) and patch each tool's configuration to route through the LLM gateway.

Token-helper tools (Claude Code, Gemini CLI) are configured to call
"thv llm token" to obtain a fresh OIDC token on demand.

Proxy-mode tools (Cursor, VS Code, Xcode) are configured to send requests to
the localhost reverse proxy started by "thv llm proxy start".

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 7 out of 7 changed files in this pull request and generated 3 comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.