Customizable Autocomplete Model Settings

[exetroncorp]

Feature Request: Customizable Autocomplete Model Settings

Summary

Add configuration options to customize the model used for autocomplete functionality, allowing users to specify alternative providers beyond the current experimental Kilo code provider.

Problem Description

Currently, the autocomplete feature is marked as "experimental" and requires the Kilo code provider, which presents several limitations:

• Corporate Environment Restrictions: Many corporate environments block or restrict access to external AI services, making the Kilo code provider inaccessible
• Limited Model Options: Users are locked into a single provider/model without flexibility to choose alternatives
• Dependency Issues: Reliance on a single external service creates potential availability and reliability concerns

Proposed Solution

Implement a settings panel or configuration option that allows users to:

1. Select Alternative Providers: Support for self-hosted models such as:

   • Codestral (self-hosted)
   • Qwen (local deployment)
   • Other compatible code completion models

2. Custom Endpoint Configuration: Allow users to specify:

   • Custom API endpoints
   • Authentication credentials (API keys, tokens)
   • Model-specific parameters

3. Provider Fallback Options: Configure primary and fallback providers for reliability

Use Cases

• Enterprise Users: Companies with strict security policies that require self-hosted solutions
• Privacy-Conscious Developers: Users who prefer to keep their code on local/private infrastructure
• Cost Optimization: Organizations wanting to use their own model deployments to control costs
• Specialized Models: Developers working with domain-specific models fine-tuned for their use cases