starred/AIClient-2-API-justlovemaki

Fork 0

mirror of https://github.com/justlovemaki/AIClient-2-API.git synced 2026-04-25 09:25:59 +03:00

[PR #174] [CLOSED] feat: Native Claude Endpoint Support & Enhanced Token Calculation #284

New issue

Closed

opened 2026-02-27 07:18:49 +03:00 by kerem · 0 comments

kerem commented

2026-02-27 07:18:49 +03:00

Owner

📋 Pull Request Information

Original PR: https://github.com/justlovemaki/AIClient-2-API/pull/174
Author: @tickernelz
Created: 1/6/2026
Status: ❌ Closed

Base: main ← Head: main

📝 Commits (4)

a5d9d89 feat: add token logging feature and enhance API request handling
ed652b7 fix: correct indentation for customNameDisplay in handleStreamRequest function
6d6c01c fix: remove unnecessary whitespace and improve code readability in toGeminiRequest method
8fc25a8 fix: remove jsonrepair dependency and fix code formatting

📊 Changes

8 files changed (+422 additions, -285 deletions)

View changed files

📝 src/api-manager.js (+8 -0)
📝 src/claude/claude-kiro.js (+358 -257)
📝 src/common.js (+8 -3)
📝 src/converters/strategies/ClaudeConverter.js (+14 -12)
📝 src/ui-manager.js (+2 -0)
📝 static/app/config-manager.js (+3 -0)
📝 static/app/i18n.js (+6 -0)
📝 static/index.html (+23 -13)

📄 Description

Summary

This update introduces full native support for Anthropic Claude API endpoints (/messages and /models), enhances token calculation accuracy for Claude Kiro, and implements a configurable token logging feature. These improvements strengthen Anthropic API compatibility and provide granular visibility into token consumption patterns.

Main Features

1. Native Claude Endpoint Support

Implemented /messages endpoint for full Anthropic API compatibility
Added /models endpoint to retrieve model information
Processes native Claude request/response formats without transformation
Maintains backward compatibility with existing /v1/messages endpoint

Implementation Files:

src/api-manager.js (lines 34-36, 59-62): Routing configuration for native endpoints
src/common.js (lines 54, 329-354): Added CLAUDE_MODEL_LIST endpoint handling logic

2. Enhanced Token Calculation for Claude Kiro

Implemented precise token accounting with detailed breakdown:
- System tokens (prompt context)
- Message tokens (text, tool interactions, media)
- Tool definition tokens (specifications and schemas)
Added token estimation for multimodal inputs:
- Images: 1600 tokens per image
- Documents: Base64 length × 0.75 / 4
Differentiated input/output token tracking during streaming

Implementation Files:

src/claude/claude-kiro.js (lines 1638-1710): Refactored estimateInputTokens() with breakdown logic
src/claude/claude-kiro.js (lines 1421-1593): Enhanced token logging in streaming responses

3. Configurable Token Logging

Introduced ENABLE_TOKEN_LOGGING configuration flag
Provides detailed token metrics for streaming and non-streaming requests
Console output includes breakdown of input, output, and tool call tokens
Configuration toggle available in application settings UI

Implementation Files:

src/claude/claude-kiro.js (lines 1165-1168, 1395-1397, 1549-1551, 1704-1706): Conditional logging implementation
static/app/config-manager.js (lines 70, 84): UI configuration handling
static/app/i18n.js: Localization strings for settings
static/index.html: Settings panel UI element

4. Model List Conversion Enhancement

Optimized model list transformation logic for native Claude endpoints
Implemented bypass for native Anthropic endpoints
Maintained OpenAI/Gemini format conversion for legacy compatibility

Implementation Files:

src/common.js (lines 347-354): Conditional conversion logic
src/converters/strategies/ClaudeConverter.js (lines 607-619, 624-637): Model list transformation strategies

5. Code Quality Improvements

Standardized formatting and whitespace consistency
Enhanced error handling with contextual logging
Improved streaming response metadata generation
Strengthened input validation at API boundaries

Technical Details

New API Endpoints

GET  /models              → Retrieves native Claude model catalog
POST /messages            → Processes messages using Anthropic API specification

Token Calculation Methodology

Input Tokens = System Tokens + Messages Tokens + Tools Tokens

Breakdown:
- System Tokens: Character-based estimation of system prompt
- Messages Tokens:
  * Text: ~4 characters per token
  * Tool interactions: Function name + JSON-serialized parameters/results
  * Images: Fixed 1600 tokens per image
  * Documents: Base64 length × 0.75 / 4
- Tools Tokens: Function name + description + JSON schema size

Configuration Parameter

{
  "ENABLE_TOKEN_LOGGING": true  // Enables detailed token consumption logging
}

Testing Recommendations

Native Endpoint Validation:
- Verify /messages endpoint compatibility with Anthropic SDK clients
- Validate /models response structure against Anthropic specification
- Confirm backward compatibility of /v1/messages endpoint
Token Accounting Verification:
- Compare estimated versus actual tokens from Anthropic API responses
- Test multimodal scenarios (images, documents, tool calls)
- Validate breakdown accuracy across message types
Conversion Compatibility:
- Validate OpenAI/Gemini model format conversion
- Test tool call parameter transformations
- Verify streaming response metadata consistency

Impact Analysis

Breaking Changes: None
Security Impact: Positive (enhanced input validation reduces injection risks)

Resolves compatibility limitations with native Anthropic API clients
Addresses token undercounting in multimodal and tool-using scenarios
Provides operational visibility into token consumption patterns

_{🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.}

## 📋 Pull Request Information **Original PR:** https://github.com/justlovemaki/AIClient-2-API/pull/174 **Author:** [@tickernelz](https://github.com/tickernelz) **Created:** 1/6/2026 **Status:** ❌ Closed **Base:** `main` ← **Head:** `main` --- ### 📝 Commits (4) - [`a5d9d89`](https://github.com/justlovemaki/AIClient-2-API/commit/a5d9d8925396fa1b140a79f7026cfd04556f525d) feat: add token logging feature and enhance API request handling - [`ed652b7`](https://github.com/justlovemaki/AIClient-2-API/commit/ed652b7939b2bf7307a3de3b0096948058ef44f4) fix: correct indentation for customNameDisplay in handleStreamRequest function - [`6d6c01c`](https://github.com/justlovemaki/AIClient-2-API/commit/6d6c01c850be4bfb6bff68f3c442cbf58c6cb362) fix: remove unnecessary whitespace and improve code readability in toGeminiRequest method - [`8fc25a8`](https://github.com/justlovemaki/AIClient-2-API/commit/8fc25a80446abe1fe47ba4e26f9d870535d6ac1a) fix: remove jsonrepair dependency and fix code formatting ### 📊 Changes **8 files changed** (+422 additions, -285 deletions) <details> <summary>View changed files</summary> 📝 `src/api-manager.js` (+8 -0) 📝 `src/claude/claude-kiro.js` (+358 -257) 📝 `src/common.js` (+8 -3) 📝 `src/converters/strategies/ClaudeConverter.js` (+14 -12) 📝 `src/ui-manager.js` (+2 -0) 📝 `static/app/config-manager.js` (+3 -0) 📝 `static/app/i18n.js` (+6 -0) 📝 `static/index.html` (+23 -13) </details> ### 📄 Description ## Summary This update introduces full native support for Anthropic Claude API endpoints (`/messages` and `/models`), enhances token calculation accuracy for Claude Kiro, and implements a configurable token logging feature. These improvements strengthen Anthropic API compatibility and provide granular visibility into token consumption patterns. ## Main Features ### 1. Native Claude Endpoint Support - Implemented `/messages` endpoint for full Anthropic API compatibility - Added `/models` endpoint to retrieve model information - Processes native Claude request/response formats without transformation - Maintains backward compatibility with existing `/v1/messages` endpoint **Implementation Files:** - `src/api-manager.js` (lines 34-36, 59-62): Routing configuration for native endpoints - `src/common.js` (lines 54, 329-354): Added `CLAUDE_MODEL_LIST` endpoint handling logic ### 2. Enhanced Token Calculation for Claude Kiro - Implemented precise token accounting with detailed breakdown: - System tokens (prompt context) - Message tokens (text, tool interactions, media) - Tool definition tokens (specifications and schemas) - Added token estimation for multimodal inputs: - Images: 1600 tokens per image - Documents: Base64 length × 0.75 / 4 - Differentiated input/output token tracking during streaming **Implementation Files:** - `src/claude/claude-kiro.js` (lines 1638-1710): Refactored `estimateInputTokens()` with breakdown logic - `src/claude/claude-kiro.js` (lines 1421-1593): Enhanced token logging in streaming responses ### 3. Configurable Token Logging - Introduced `ENABLE_TOKEN_LOGGING` configuration flag - Provides detailed token metrics for streaming and non-streaming requests - Console output includes breakdown of input, output, and tool call tokens - Configuration toggle available in application settings UI **Implementation Files:** - `src/claude/claude-kiro.js` (lines 1165-1168, 1395-1397, 1549-1551, 1704-1706): Conditional logging implementation - `static/app/config-manager.js` (lines 70, 84): UI configuration handling - `static/app/i18n.js`: Localization strings for settings - `static/index.html`: Settings panel UI element ### 4. Model List Conversion Enhancement - Optimized model list transformation logic for native Claude endpoints - Implemented bypass for native Anthropic endpoints - Maintained OpenAI/Gemini format conversion for legacy compatibility **Implementation Files:** - `src/common.js` (lines 347-354): Conditional conversion logic - `src/converters/strategies/ClaudeConverter.js` (lines 607-619, 624-637): Model list transformation strategies ### 5. Code Quality Improvements - Standardized formatting and whitespace consistency - Enhanced error handling with contextual logging - Improved streaming response metadata generation - Strengthened input validation at API boundaries ## Technical Details ### New API Endpoints ``` GET /models → Retrieves native Claude model catalog POST /messages → Processes messages using Anthropic API specification ``` ### Token Calculation Methodology ``` Input Tokens = System Tokens + Messages Tokens + Tools Tokens Breakdown: - System Tokens: Character-based estimation of system prompt - Messages Tokens: * Text: ~4 characters per token * Tool interactions: Function name + JSON-serialized parameters/results * Images: Fixed 1600 tokens per image * Documents: Base64 length × 0.75 / 4 - Tools Tokens: Function name + description + JSON schema size ``` ### Configuration Parameter ```json { "ENABLE_TOKEN_LOGGING": true // Enables detailed token consumption logging } ``` ## Testing Recommendations 1. **Native Endpoint Validation:** - Verify `/messages` endpoint compatibility with Anthropic SDK clients - Validate `/models` response structure against Anthropic specification - Confirm backward compatibility of `/v1/messages` endpoint 2. **Token Accounting Verification:** - Compare estimated versus actual tokens from Anthropic API responses - Test multimodal scenarios (images, documents, tool calls) - Validate breakdown accuracy across message types 3. **Conversion Compatibility:** - Validate OpenAI/Gemini model format conversion - Test tool call parameter transformations - Verify streaming response metadata consistency ## Impact Analysis - **Breaking Changes:** None - **Security Impact:** Positive (enhanced input validation reduces injection risks) ## Related Issues - Resolves compatibility limitations with native Anthropic API clients - Addresses token undercounting in multimodal and tool-using scenarios - Provides operational visibility into token consumption patterns --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

kerem

2026-02-27 07:18:49 +03:00

closed this issue
added the
pull-request
label

No labels

pull-request

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference

starred/AIClient-2-API-justlovemaki#284

No description provided.

Rows
Columns