Basic project scaffold with: chat completion, streaming and models endpoints from OpenAI

[SantiagoDePolonia]

No description provided.

[Copilot]

Pull request overview

This PR establishes the foundational scaffold for HeavyModel, a high-performance LLM gateway written in Go. It provides OpenAI-compatible endpoints for chat completions (both streaming and non-streaming) and model listing, with a clean provider abstraction to support multiple LLM backends.

Key Changes:

• Implements core provider interface with OpenAI integration supporting GPT and O1 model families
• Sets up HTTP server with Echo framework providing /v1/chat/completions, /v1/models, and /health endpoints
• Adds configuration management via Viper with support for environment variables and YAML config files

Reviewed changes

Copilot reviewed 11 out of 13 changed files in this pull request and generated 9 comments.

Show a summary per file

File	Description
go.mod, go.sum	Defines Go 1.21 module with dependencies for Echo web framework and Viper configuration
Makefile	Provides build, run, test, and clean targets for development workflow
.gitignore	Basic ignore rules for compiled binary (minimal coverage)
README.md	Quick start documentation with Docker support and endpoint descriptions
config/config.yaml	Server configuration with port 8088 default and OpenAI API key placeholder
config/config.go	Configuration loader with Viper integration and environment variable override support
internal/core/types.go	Defines request/response types for chat completions and model listings
internal/core/interfaces.go	Provider interface defining model support checks, completions, and streaming
internal/providers/openai/openai.go	OpenAI provider implementation with HTTP client for API communication
internal/server/http.go	HTTP server setup with Echo, middleware, and route registration
internal/server/handlers.go	Request handlers for chat completions, streaming, health checks, and model listing
internal/server/handlers_test.go	Test suite covering successful request flows and basic error cases

---

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

[Copilot]

Exposing internal error details directly to API clients can leak sensitive information about the system's implementation. The error from h.provider.StreamChatCompletion could contain internal details, API keys, or stack traces. Consider using a generic error message for the client while logging the detailed error internally:

return c.JSON(http.StatusInternalServerError, map[string]string{"error": "failed to process streaming chat completion"})

[Copilot]

Exposing internal error details directly to API clients can leak sensitive information. The error from h.provider.ListModels could contain internal details or API keys. Consider using a generic error message for the client while logging the detailed error internally:

return c.JSON(http.StatusInternalServerError, map[string]string{"error": "failed to retrieve models"})

[Copilot]

The HTTP server lacks timeout configurations, which could lead to resource exhaustion from slow or hanging connections. Consider configuring timeouts when starting the server:

func (s *Server) Start(addr string) error {
    server := &http.Server{
        Addr:         addr,
        Handler:      s.echo,
        ReadTimeout:  15 * time.Second,
        WriteTimeout: 15 * time.Second,
        IdleTimeout:  60 * time.Second,
    }
    return server.ListenAndServe()
}

Note: You'll need to import "net/http" and "time".

[Copilot]

Missing input validation for the chat request. The handler should validate that:

• The Messages array is not empty
• The Model field is not an empty string
• Message roles and content are not empty

Consider adding validation before processing the request:

if req.Model == "" {
    return c.JSON(http.StatusBadRequest, map[string]string{
        "error": "model is required",
    })
}
if len(req.Messages) == 0 {
    return c.JSON(http.StatusBadRequest, map[string]string{
        "error": "messages array cannot be empty",
    })
}

[Copilot]

The io.Copy call does not check for errors. If copying the stream fails, the error is silently ignored. Consider checking and logging the error:

if _, err := io.Copy(c.Response().Writer, stream); err != nil {
    // Log error or handle appropriately
}

[Copilot]

The model prefix "o1" should include a hyphen to properly match OpenAI's o1 model naming (e.g., "o1-preview", "o1-mini"). The current check would incorrectly match models starting with "o1" but not hyphenated, like "o123model". Consider using:

return strings.HasPrefix(model, "gpt-") || strings.HasPrefix(model, "o1-")

Suggested change

	return strings.HasPrefix(model, "gpt-") || strings.HasPrefix(model, "o1")
	return strings.HasPrefix(model, "gpt-") || strings.HasPrefix(model, "o1-")

[Copilot]

Exposing internal error details directly to API clients can leak sensitive information about the system's implementation. The error from h.provider.ChatCompletion could contain internal details, API keys, or stack traces. Consider using a generic error message for the client while logging the detailed error internally:

return c.JSON(http.StatusInternalServerError, map[string]string{"error": "failed to process chat completion"})

[Copilot]

Missing test coverage for error scenarios in the ChatCompletion handler. While TestListModelsError demonstrates error handling for the ListModels endpoint, there are no tests for error cases in ChatCompletion, such as:

• Provider returning an error during non-streaming completion
• Provider returning an error during streaming completion
• Invalid request body (malformed JSON)

Consider adding test cases to cover these error paths.

[Copilot]

The golang.org/x/crypto version v0.17.0 has known vulnerabilities. Consider upgrading to at least v0.31.0 or later to address security issues. You can update by running:

go get golang.org/x/crypto@latest

Suggested change

	golang.org/x/crypto v0.17.0 // indirect
	golang.org/x/crypto v0.31.0 // indirect

[SantiagoDePolonia]

The validation should be moved to config loader