Add Kiro CLI provider with multi-credential support by ItzMeAyben · Pull Request #116 · Mirrowel/LLM-API-Key-Proxy

ItzMeAyben · 2026-02-02T18:48:32Z

Summary

Add Kiro CLI (Amazon Q / Kiro) provider enabling access to Claude models via AWS Bedrock using Kiro CLI authentication.

Port Note: Uses 7777 (project default) vs reference's 8000.

Features

Multi-credential authentication - Environment variables, JSON files, SQLite database with auto-detection
Automatic token refresh - Refreshes expired tokens via AWS SSO OIDC/Kiro Desktop endpoints automatically
Token persistence - Saves refreshed tokens back to source (JSON/SQLite)
Dual authentication types - Kiro Desktop and AWS SSO OIDC (auto-detected)
AWS Event Stream parsing - Proper handling of streaming responses with <thinking> block extraction
Full OpenAI compatibility - Chat completions, streaming, tool calling, vision support
Cross-platform - macOS and Linux path support

Supported Models

All Claude models available through Kiro API:

Model	Thinking Support	Context Limit
claude-sonnet-4-5	✅	200K
claude-opus-4-5	✅	200K
claude-sonnet-4	✅	200K
claude-haiku-4-5	✅	200K
claude-3-7-sonnet	✅	200K

Architecture

Authentication Flow

sequenceDiagram
    participant Client as API Client
    participant Proxy as LLM Proxy
    participant Auth as KiroAuthManager
    participant Store as Credential Store
    participant OIDC as AWS SSO/Kiro Auth
    participant API as Kiro Bedrock API

    Client->>Proxy: POST /v1/chat/completions
    Proxy->>Auth: get_access_token()
    Auth->>Store: Load credentials
    Store-->>Auth: Token + metadata
    
    alt Token valid (> 10min remaining)
        Auth-->>Proxy: Valid access_token
    else Token expired or expiring
        Auth->>OIDC: Refresh token request
        OIDC-->>Auth: New access_token (1hr)
        Auth->>Store: Save refreshed token
        Auth-->>Proxy: New access_token
    end
    
    Proxy->>API: Request with Bearer token
    API-->>Proxy: Event Stream response
    Proxy-->>Client: OpenAI SSE chunks

Token Management

Automatic Refresh

Access tokens expire in 1 hour
Tokens automatically refresh 10 minutes before expiry
Refreshed tokens saved back to credential source
Supports both Kiro Desktop and AWS SSO OIDC flows

Token Storage Locations

OS	Path
macOS	`~/Library/Application Support/kiro-cli/data.sqlite3`
macOS	`~/.aws/sso/cache/kiro-auth-token.json`
Linux	`~/.local/share/kiro-cli/data.sqlite3`
Linux	`~/.aws/sso/cache/kiro-auth-token.json`

Authentication Priority

KIRO_REFRESH_TOKEN / REFRESH_TOKEN environment variable
JSON credentials file (KIRO_CREDS_FILE)
SQLite database (KIRO_CLI_DB_FILE)
Auto-detected default paths

System Architecture

graph TB
    subgraph "LLM Proxy Server"
        A[FastAPI Router] --> B[KiroCliProvider]
        B --> C[Message Converters]
        B --> D[Stream Parser]
        B --> E[HTTP Client]
    end
    
    subgraph "Authentication Layer"
        F[KiroAuthManager] --> G[(JSON Files)]
        F --> H[(SQLite DB)]
        F --> I[Token Refresh]
        I --> J[AWS SSO OIDC]
        I --> K[Kiro Desktop Auth]
    end
    
    subgraph "External Services"
        L[Kiro Bedrock API]
    end
    
    B --> F
    E -->|Bearer Token| L
    F -->|Credentials| E
    L -->|AWS Event Stream| D
    D -->|OpenAI SSE| A
    C -->|Kiro Payload| E

Project Structure

src/
└── rotator_library/                    # Core library (LGPL-3.0-only)
    ├── credential_manager.py           # Modified: Kiro credential discovery
    └── providers/                      # LLM provider implementations
        ├── kiro_auth_base.py           # NEW: Token lifecycle & auth manager
        ├── kiro_cli_provider.py        # NEW: Provider implementation
        └── utilities/
            ├── kiro_converters.py      # NEW: Message/tool format conversion
            ├── kiro_http_client.py     # NEW: HTTP client with retry logic
            ├── kiro_streaming.py       # NEW: AWS Event Stream parser
            └── kiro_utils.py           # NEW: Configuration & utilities

src/proxy_app/                          # Proxy application (MIT)
├── launcher_tui.py                     # Modified: TUI credential detection
└── main.py                             # Modified: Skip JSON parsing for SQLite

docker/
├── docker-compose.dev.yml              # Modified: Kiro credential volume mounts
└── .env.example                        # Modified: Kiro auth documentation

Testing

Verified working across multiple scenarios:

✅ JSON credentials (Kiro IDE, macOS)
✅ SQLite credentials (kiro-cli, macOS)
✅ Environment variable authentication
✅ Auto-detection across credential types
✅ Token refresh and persistence
✅ Chat completions endpoint
✅ Streaming responses with thinking blocks
✅ Tool calling support
✅ Vision support
✅ Multi-turn conversations
✅ Model listing
✅ Docker deployment with volume mounts
✅ Cross-platform paths (macOS/Linux)

Implementation Notes

Provider Interface: Follows ProviderInterface pattern from provider_interface.py
Custom Logic: Implements has_custom_logic() for direct API handling
Licensing: LGPL-3.0 headers on all new files (per CONTRIBUTING.md)
Integration: Seamlessly integrated with existing credential system
Documentation: Comprehensive setup guide in README.md & .env.example
Error Handling: Exponential backoff for 403/429/5xx with automatic token refresh on 403

Checklist

Code follows project's style guidelines
LGPL-3.0 license headers added to all new files
Follows ProviderInterface pattern
Integrated with existing credential management
Cross-platform support (macOS/Linux)
Documentation added (README.md, .env.example)
Docker deployment support
Tested with multiple authentication methods

Acknowledgments

Implementation adapted from kiro-gateway by @Jwadow - reference for Kiro API flows, AWS Event Stream parsing, and message format conversions. Integrated into LLM-API-Key-Proxy's provider architecture with LGPL-3.0 licensing.

…and documentation

Add comprehensive Kiro CLI (Amazon Q / Kiro) provider implementation with support for multiple authentication methods: - Direct refresh token via environment variables - JSON credentials from Kiro IDE (~/.aws/sso/cache/kiro-auth-token.json) - SQLite database from kiro-cli (macOS and Linux) - Auto-detection with configurable priority Features: - KiroAuthManager with AWS SSO OIDC and Kiro Desktop auth support - Token refresh with automatic expiration handling - HTTP client with retry logic (403, 429, 5xx) - OpenAI message/tool format conversion to Kiro API - AWS Event Stream parsing for streaming responses - Thinking block extraction and formatting - macOS and Linux path support Updates: - credential_manager.py: Add Kiro credential discovery - launcher_tui.py: Add Kiro credential detection in TUI - main.py: Skip JSON parsing for Kiro SQLite credentials - docker-compose.dev.yml: Add Kiro credential volume mounts - .env.example: Document all Kiro auth options - README.md: Add comprehensive Kiro setup documentation Co-authored-by: Cursor <cursoragent@cursor.com>

Add SPDX-License-Identifier and Copyright headers to all new Kiro provider files as required by CONTRIBUTING.md guidelines. Co-authored-by: Cursor <cursoragent@cursor.com>

Mirrowel · 2026-03-03T10:58:40Z

@greptile

Mirrowel · 2026-03-03T11:00:05Z

@ItzMeAyben I didn't notice before, but the base branch here is main - it should be Dev branch instead. Main is never directly modified and is outdated.

greptile-apps · 2026-03-03T11:02:29Z

Greptile Summary

This PR adds comprehensive Kiro CLI provider support, enabling access to Claude models via AWS Bedrock using Kiro CLI/IDE authentication. The implementation is well-architected and follows the existing provider interface pattern.

Key Changes:

Implemented complete authentication system supporting three credential sources: environment variables, JSON files (Kiro IDE), and SQLite databases (kiro-cli)
Built automatic credential refresh with 10-minute safety margin, supporting both Kiro Desktop and AWS SSO OIDC auth flows
Created AWS Event Stream parser with thinking block extraction and OpenAI SSE conversion
Added message/tool format converters with schema sanitization for API compatibility
Updated default port from 8000 to 7777 across all configurations and documentation
Integrated credential auto-detection into existing discovery system

Implementation Quality:

All new files include proper LGPL-3.0 license headers as required
Follows established ProviderInterface pattern with custom logic support
Comprehensive error handling with exponential backoff and automatic retry
Thread-safe token refresh using asyncio locks to prevent race conditions
Cross-platform support with proper path handling for macOS and Linux
Extensive documentation in README and .env.example

Testing Coverage:
The PR description indicates thorough testing across multiple authentication methods, API features (streaming, tools, vision), and deployment scenarios (local and Docker).

Minor Observations:

Two minor style improvements suggested in inline comments
Otherwise, code quality is high with proper error handling, parameterized SQL queries, and defensive programming

Confidence Score: 5/5

This PR is safe to merge - well-structured implementation with proper error handling and comprehensive testing
Score reflects excellent code quality: follows established patterns, includes proper licensing, has comprehensive error handling, uses parameterized SQL queries (no injection risk), implements thread-safe token refresh, and includes extensive documentation. Only minor style suggestions identified.
No files require special attention - all implementations follow best practices

Important Files Changed

Filename	Overview
src/rotator_library/providers/kiro_auth_base.py	Core authentication manager with token lifecycle, credential loading (JSON/SQLite), automatic refresh, and multi-auth support (Kiro Desktop/AWS SSO OIDC)
src/rotator_library/providers/kiro_cli_provider.py	Provider implementation following ProviderInterface pattern, handles model resolution, dynamic model discovery, and streaming/non-streaming responses
src/rotator_library/providers/utilities/kiro_converters.py	Message format conversions between OpenAI and Kiro APIs, handles tools, images, thinking tags injection, and schema sanitization for compatibility
src/rotator_library/providers/utilities/kiro_streaming.py	AWS Event Stream parser with thinking block extraction, tool call parsing, OpenAI SSE conversion, handles first token timeout and retry logic
src/rotator_library/credential_manager.py	Added Kiro CLI credential discovery supporting env tokens, JSON files, and SQLite databases with priority-based detection across platforms
src/proxy_app/main.py	Changed default port from 8000 to 7777, added special handling to skip JSON parsing for kiro_cli SQLite credentials during initialization
README.md	Added comprehensive Kiro CLI documentation covering three authentication methods, auto-detection paths, configuration options, and examples

Sequence Diagram

sequenceDiagram
    participant Client as API Client
    participant Proxy as LLM Proxy
    participant Provider as KiroCliProvider
    participant Auth as KiroAuthManager
    participant Store as Credential Store<br/>(JSON/SQLite)
    participant OIDC as AWS SSO/Kiro Auth
    participant API as Kiro Bedrock API

    Client->>Proxy: POST /v1/chat/completions<br/>model: kiro_cli/claude-sonnet-4.5
    Proxy->>Provider: acompletion(messages, tools)
    Provider->>Auth: Request authentication
    
    alt Credentials valid (more than 10min remaining)
        Auth-->>Provider: Return valid auth header
    else Credentials expired or expiring soon
        Auth->>Store: Reload from disk
        Store-->>Auth: Latest credentials with metadata
        
        alt Still expired after reload
            Auth->>OIDC: POST refresh request
            OIDC-->>Auth: New credentials with expiry
            Auth->>Store: Persist updated credentials
        end
        
        Auth-->>Provider: Return fresh auth header
    end
    
    Provider->>Provider: Convert OpenAI to Kiro format
    Provider->>API: POST /generateAssistantResponse<br/>with Bearer authorization
    
    alt Streaming enabled
        API-->>Provider: AWS Event Stream
        Provider->>Provider: Parse events, extract thinking
        Provider-->>Client: OpenAI SSE chunks
    else Non-streaming
        API-->>Provider: AWS Event Stream
        Provider->>Provider: Collect full response
        Provider-->>Client: OpenAI completion
    end

_{Last reviewed commit: 076746c}

greptile-apps

_{23 files reviewed, 2 comments}

_{Edit Code Review Agent Settings | Greptile}

greptile-apps · 2026-03-03T11:02:33Z

src/rotator_library/providers/utilities/kiro_http_client.py

+                    await self.auth_manager.force_refresh()
+                    continue
+
+                if response.status_code in (429,) or 500 <= response.status_code < 600:


Single-element tuple (429,) is unnecessarily verbose

Suggested change

if response.status_code in (429,) or 500 <= response.status_code < 600:

if response.status_code == 429 or 500 <= response.status_code < 600:

Prompt To Fix With AI

This is a comment left during a code review. Path: src/rotator_library/providers/utilities/kiro_http_client.py Line: 96 Comment: Single-element tuple `(429,)` is unnecessarily verbose ```suggestion if response.status_code == 429 or 500 <= response.status_code < 600: ``` How can I resolve this? If you propose a fix, please make it concise.

_{Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!}

greptile-apps · 2026-03-03T11:02:34Z

src/rotator_library/credential_manager.py


+        # Special-case: Kiro CLI supports multiple credential sources
+        # Priority: REFRESH_TOKEN (env) > KIRO_CREDS_FILE (JSON) > KIRO_CLI_DB_FILE (SQLite) > default paths
+        kiro_refresh_token = self.env_vars.get("KIRO_REFRESH_TOKEN") or self.env_vars.get("REFRESH_TOKEN")


Consider using only KIRO_REFRESH_TOKEN instead of falling back to generic REFRESH_TOKEN to avoid potential conflicts with other systems

Prompt To Fix With AI

This is a comment left during a code review. Path: src/rotator_library/credential_manager.py Line: 129 Comment: Consider using only `KIRO_REFRESH_TOKEN` instead of falling back to generic `REFRESH_TOKEN` to avoid potential conflicts with other systems How can I resolve this? If you propose a fix, please make it concise.

_{Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!}

mirrobot-agent · 2026-03-03T11:10:53Z

Starting my review of this Kiro CLI provider implementation. Adding support for Claude models via AWS Bedrock with multi-credential authentication - this looks comprehensive! Analyzing the new provider files now. 🔍

mirrobot-agent

Overall Assessment

Excellent PR implementing comprehensive Kiro CLI provider support! The architecture is well-designed with clean separation of concerns across authentication, HTTP client, streaming, and conversion layers. The code follows the existing provider interface pattern and includes proper LGPL-3.0 licensing on all new files.

Key Strengths

Multi-credential authentication: Clean implementation supporting environment variables, JSON files, and SQLite databases with auto-detection
Token lifecycle management: Proper refresh handling with 10-minute safety margin and persistent storage
AWS Event Stream parsing: Well-implemented parser for streaming responses with thinking block extraction
Error handling: Good use of exponential backoff and automatic token refresh on 403 errors
Thread safety: Proper use of asyncio.Lock for token refresh to prevent race conditions

Notes

I concur with Greptile's observations regarding:

The verbose tuple syntax in kiro_http_client.py:96 could be simplified
The REFRESH_TOKEN fallback in credential_manager.py:129 could conflict with other systems

Minor Suggestion

The silent JSON parsing failure in extract_tool_uses_from_message() could mask issues with malformed tool arguments. Consider adding logging when json.loads() fails.

Questions for the Author

Has this been tested with both Kiro Desktop and AWS SSO OIDC authentication flows in production?
Are there any plans to support Windows paths for the SQLite auto-detection?

Overall, this is high-quality code ready for merge after addressing the minor suggestions.

This review was generated by an AI assistant.

mirrobot-agent · 2026-03-03T11:12:41Z

src/rotator_library/providers/utilities/kiro_converters.py

+            func = tc.get("function", {})
+            arguments = func.get("arguments", "{}")
+            if isinstance(arguments, str):


If json.loads() fails here, input_data becomes an empty dict but the code continues silently. Consider logging this parsing failure so malformed tool arguments don't go unnoticed.

ItzMeAyben and others added 3 commits February 2, 2026 23:04

fix: 🐛 update proxy port from 8000 to 7777 across all configurations …

ce3b188

…and documentation

chore: Add required LGPL-3.0 license headers to Kiro provider files

076746c

Add SPDX-License-Identifier and Copyright headers to all new Kiro provider files as required by CONTRIBUTING.md guidelines. Co-authored-by: Cursor <cursoragent@cursor.com>

ItzMeAyben requested a review from Mirrowel as a code owner February 2, 2026 18:48

Repository owner deleted a comment from mirrobot-agent bot Mar 3, 2026

Mirrowel added enhancement New feature or request Agent Monitored Monitored for AI Agent to review PR's and commits Priority labels Mar 3, 2026

greptile-apps bot reviewed Mar 3, 2026

View reviewed changes

mirrobot-agent bot reviewed Mar 3, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Add Kiro CLI provider with multi-credential support#116

Add Kiro CLI provider with multi-credential support#116
ItzMeAyben wants to merge 3 commits intoMirrowel:mainfrom
ItzMeAyben:feature/kiro-cli-provider

ItzMeAyben commented Feb 2, 2026

Uh oh!

Mirrowel commented Mar 3, 2026

Uh oh!

Mirrowel commented Mar 3, 2026

Uh oh!

greptile-apps bot commented Mar 3, 2026

Greptile Summary

Confidence Score: 5/5

Important Files Changed

Sequence Diagram

Uh oh!

greptile-apps bot left a comment

Uh oh!

greptile-apps bot Mar 3, 2026

Uh oh!

greptile-apps bot Mar 3, 2026

Uh oh!

mirrobot-agent bot commented Mar 3, 2026

Uh oh!

mirrobot-agent bot left a comment

Uh oh!

mirrobot-agent bot Mar 3, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

	if response.status_code in (429,) or 500 <= response.status_code < 600:
	if response.status_code == 429 or 500 <= response.status_code < 600:

Uh oh!

Conversation

ItzMeAyben commented Feb 2, 2026

Summary

Features

Supported Models

Architecture

Authentication Flow

Token Management

System Architecture

Project Structure

Testing

Implementation Notes

Checklist

Acknowledgments

Uh oh!

Mirrowel commented Mar 3, 2026

Uh oh!

Mirrowel commented Mar 3, 2026

Uh oh!

greptile-apps bot commented Mar 3, 2026

Greptile Summary

Confidence Score: 5/5

Important Files Changed

Sequence Diagram

Uh oh!

greptile-apps bot left a comment

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot Mar 3, 2026

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot Mar 3, 2026

Choose a reason for hiding this comment

Uh oh!

mirrobot-agent bot commented Mar 3, 2026

Uh oh!

mirrobot-agent bot left a comment

Choose a reason for hiding this comment

Overall Assessment

Key Strengths

Notes

Minor Suggestion

Questions for the Author

Uh oh!

mirrobot-agent bot Mar 3, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants