Feat/add litellm provider

[RheagalFire]

Summary

• Adds LiteLLM as a first-class AI gateway provider, giving CodeGPT users access to 100+ LLM providers (Anthropic, Bedrock, Vertex AI, Groq, Mistral, Cohere, etc.) through a single proxy endpoint.
• No new Go dependencies - reuses the existing go-openai SDK since the LiteLLM proxy speaks the OpenAI Chat Completions protocol.

Motivation

CodeGPT currently supports 4 providers (OpenAI, Azure, Gemini, Anthropic) plus OpenAI-compatible endpoints via base_url. Users who want to use providers like AWS Bedrock, Google Vertex AI, Groq, Mistral, or Cohere need to set up separate proxy infrastructure. LiteLLM is a widely-adopted open-source AI gateway (45k+ GitHub stars) that normalizes 100+ provider APIs into the OpenAI format - users run litellm --model anthropic/claude-sonnet-4-6 and get a local OpenAI-compatible server at localhost:4000.

Adding LiteLLM as a named provider (instead of requiring users to manually configure base_url) provides:

• Dedicated litellm.* config namespace (api_key, base_url, api_key_helper)
• Secure credential storage via the existing keyring system
• Clear provider selection: codegpt config set openai.provider litellm

Changes

• provider/litellm/litellm.go - New LiteLLM client implementing core.Generative (Completion, CompletionStream, GetSummaryPrefix with function-call fallback)
• provider/litellm/options.go - Functional options pattern matching existing providers (WithToken, WithModel, WithBaseURL, etc.)
• provider/litellm/func.go - Conventional commit prefix extraction via OpenAI-format tool calls
• provider/litellm/func_test.go - Unit test for summary prefix argument parsing
• provider/litellm/options_test.go - Config validation, defaults, and option application tests
• provider/litellm/stream_test.go - Mock SSE server tests for streaming, completion, function calling, fallback, and error handling
• core/platform.go - Added LiteLLM platform constant
• cmd/provider.go - Added NewLiteLLM() factory and GetClient() dispatch
• cmd/config_list.go - Added litellm.* config key descriptions
• cmd/config_set.go - Added litellm.* CLI flags and viper bindings
• cmd/cmd.go - Added litellm.api_key to sensitive credential store

Tests

1. Unit tests (12 tests, 89.9% coverage):

=== RUN   TestGetSummaryPrefixArgs
--- PASS: TestGetSummaryPrefixArgs (0.00s)
=== RUN   Test_config_valid
    --- PASS: Test_config_valid/valid_config (0.00s)
    --- PASS: Test_config_valid/missing_token (0.00s)
    --- PASS: Test_config_valid/missing_model (0.00s)
=== RUN   Test_config_defaults
--- PASS: Test_config_defaults (0.00s)
=== RUN   Test_config_options
--- PASS: Test_config_options (0.00s)
=== RUN   TestNew_missingToken
--- PASS: TestNew_missingToken (0.00s)
=== RUN   TestNew_missingModel
--- PASS: TestNew_missingModel (0.00s)
=== RUN   TestNew_success
--- PASS: TestNew_success (0.00s)
=== RUN   TestCompletionStream
--- PASS: TestCompletionStream (0.00s)
=== RUN   TestCompletion
--- PASS: TestCompletion (0.00s)
=== RUN   TestGetSummaryPrefix
--- PASS: TestGetSummaryPrefix (0.00s)
=== RUN   TestGetSummaryPrefix_fallback
--- PASS: TestGetSummaryPrefix_fallback (0.00s)
=== RUN   TestCompletion_noChoices
--- PASS: TestCompletion_noChoices (0.00s)
PASS
ok  	github.com/appleboy/CodeGPT/provider/litellm	coverage: 89.9% of statements

2. Full test suite - 0 regressions across all packages.

3. Lint clean: make lint -> 0 issues.

4. Build clean: go build ./... succeeds with no errors.

Risk / Compatibility

• Additive only - existing providers (OpenAI, Azure, Gemini, Anthropic) are untouched.
• No new Go dependencies - reuses the existing sashabaranov/go-openai SDK.
• LiteLLM API key is stored in the secure credential store (keyring), same as other providers.
• Falls back from litellm.api_key to openai.api_key if not set.

Example usage

# 1. Start a LiteLLM proxy (separate terminal)
litellm --model anthropic/claude-sonnet-4-6

# 2. Configure CodeGPT to use LiteLLM
codegpt config set openai.provider litellm
codegpt config set litellm.api_key sk-your-litellm-key
codegpt config set openai.model anthropic/claude-sonnet-4-6

# Optional: custom proxy URL (default: http://localhost:4000/v1)
codegpt config set litellm.base_url http://my-litellm-server:4000/v1

# 3. Use as normal
codegpt commit --preview
codegpt review

[codecov]

Codecov Report

❌ Patch coverage is 78.02198% with 60 lines in your changes missing coverage. Please review.
✅ Project coverage is 32.90%. Comparing base (0125d86) to head (9510e84).
⚠️ Report is 297 commits behind head on main.

Files with missing lines	Patch %	Lines
cmd/provider.go	0.00%	52 Missing ⚠️
provider/litellm/litellm.go	96.77%	4 Missing ⚠️
provider/litellm/options.go	97.26%	2 Missing ⚠️
core/platform.go	0.00%	1 Missing ⚠️
provider/litellm/func.go	83.33%	1 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##             main     #268       +/-   ##
===========================================
+ Coverage   21.96%   32.90%   +10.94%     
===========================================
  Files           8       39       +31     
  Lines         346     2194     +1848     
===========================================
+ Hits           76      722      +646     
- Misses        264     1471     +1207     
+ Partials        6        1        -5     

Flag	Coverage Δ
ubuntu-latest	32.90% <78.02%> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 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.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[appleboy]

I will take it.