--- title: http_pool | CodeWeaver Docs description: API reference for codeweaver.providers.http_pool url: "https://docs.knitli.com/api/providers/http_pool" type: static generatedAt: "2026-04-17T17:21:09.259Z" --- # http_pool [Open in ChatGPT](https://chatgpt.com/?q=Read%20https%3A%2F%2Fdocs.knitli.com%2Fcodeweaver%2Fapi%2Fproviders%2Fhttp_pool%2F.%20I%20want%20to%20ask%20questions%20about%20it.)[Open in Claude](https://claude.ai/new?q=Read%20https%3A%2F%2Fdocs.knitli.com%2Fcodeweaver%2Fapi%2Fproviders%2Fhttp_pool%2F.%20I%20want%20to%20ask%20questions%20about%20it.)[View in Markdown](/codeweaver/api/providers/http_pool.md) [Share on LinkedIn](https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fdocs.knitli.com%2Fcodeweaver%2Fapi%2Fproviders%2Fhttp_pool%2F)[Share on X](https://x.com/intent/tweet?url=https%3A%2F%2Fdocs.knitli.com%2Fcodeweaver%2Fapi%2Fproviders%2Fhttp_pool%2F&text=http_pool)[Share on Threads](https://threads.net/intent/post?url=https%3A%2F%2Fdocs.knitli.com%2Fcodeweaver%2Fapi%2Fproviders%2Fhttp_pool%2F&text=http_pool)[Share on Bluesky](https://bsky.app/intent/compose?text=http_pool%20https%3A%2F%2Fdocs.knitli.com%2Fcodeweaver%2Fapi%2Fproviders%2Fhttp_pool%2F)[Share on Facebook](https://www.facebook.com/sharer/sharer.php?u=https%3A%2F%2Fdocs.knitli.com%2Fcodeweaver%2Fapi%2Fproviders%2Fhttp_pool%2F)[Share on Reddit](https://reddit.com/submit?url=https%3A%2F%2Fdocs.knitli.com%2Fcodeweaver%2Fapi%2Fproviders%2Fhttp_pool%2F&title=http_pool)[Share on Hacker News](https://news.ycombinator.com/submitlink?u=https%3A%2F%2Fdocs.knitli.com%2Fcodeweaver%2Fapi%2Fproviders%2Fhttp_pool%2F&t=http_pool)[Share on Email](mailto:?subject=http_pool&body=https%3A%2F%2Fdocs.knitli.com%2Fcodeweaver%2Fapi%2Fproviders%2Fhttp_pool%2F)[Share on WhatsApp](https://wa.me/?text=http_pool%20https%3A%2F%2Fdocs.knitli.com%2Fcodeweaver%2Fapi%2Fproviders%2Fhttp_pool%2F)[Share on Telegram](https://t.me/share/url?url=https%3A%2F%2Fdocs.knitli.com%2Fcodeweaver%2Fapi%2Fproviders%2Fhttp_pool%2F&text=http_pool) # `codeweaver.providers.http_pool` [Section titled “codeweaver.providers.http_pool”](#codeweaverprovidershttp_pool) HTTP client connection pool manager for CodeWeaver providers. This module provides centralized HTTP client pooling for providers that use httpx, including Voyage AI, Cohere, and other HTTP-based API providers. Connection pooling reduces overhead from repeated TCP handshakes and TLS negotiations, improving performance and reliability during high-load operations like indexing. Usage: from codeweaver.core.http_pool import HttpClientPool, get_http_pool # Get singleton instance [Section titled “Get singleton instance”](#get-singleton-instance) pool = get_http_pool() # Get a pooled client for a specific provider [Section titled “Get a pooled client for a specific provider”](#get-a-pooled-client-for-a-specific-provider) client = await pool.get_client(‘voyage’, read_timeout=90.0) # Cleanup on shutdown [Section titled “Cleanup on shutdown”](#cleanup-on-shutdown) await pool.close_all() Thread Safety: - Singleton initialization: Thread-safe via double-checked locking (threading.Lock) - Async client creation (get_client): Coroutine-safe via asyncio.Lock - Sync client creation (get_client_sync): Thread-safe via threading.Lock - WARNING: Synchronous and asynchronous methods should NOT be mixed. The locking strategy does not protect against concurrent access from both threads and coroutines. Only use sync methods in a purely threaded context, and async methods in a purely async context. - All methods can be called concurrently within their respective contexts (sync or async) without creating duplicate clients. ## Class: `HttpClientPool` [Section titled “Class: HttpClientPool”](#class-httpclientpool) Singleton HTTP client pool manager for provider connections. Manages a collection of httpx.AsyncClient instances, one per provider, with configurable connection limits and timeouts. Clients are created lazily on first request and reused for subsequent requests. The pool supports HTTP/2 by default for better multiplexing on modern APIs. Thread Safety: Client creation is protected by an asyncio.Lock to prevent race conditions when multiple coroutines request the same client simultaneously. Example: pool = HttpClientPool.get_instance() client = await pool.get_client(‘voyage’, max_connections=50, read_timeout=90.0) # Use client for API calls… [Section titled “Use client for API calls…”](#use-client-for-api-calls) await pool.close_all() # Cleanup on shutdown ### Method: `close_all` [Section titled “Method: close_all”](#method-close_all) ** ``` close_all() ``` Close all pooled clients concurrently (cleanup on shutdown). All clients are closed in parallel to minimize shutdown latency. Expected errors (HTTPError, OSError) are logged as warnings. Unexpected exceptions are re-raised after all close operations complete. This method should be called during application shutdown to properly close all HTTP connections and release resources. ### Method: `close_client` [Section titled “Method: close_client”](#method-close_client) ** ``` close_client() ``` Close a specific provider’s HTTP client. Args: name: Provider name whose client to close. Returns: True if client was closed, False if no client existed. ### Method: `get_client` [Section titled “Method: get_client”](#method-get_client) ** ``` get_client() ``` Get or create a pooled HTTP client for a specific provider. Clients are cached by name and reused for subsequent requests. Override parameters allow per-provider customization of limits and timeouts. This method is coroutine-safe: concurrent calls for the same provider will not create duplicate clients. Note: Clients are cached by name only, not by override parameters. If the same provider name is requested multiple times with different overrides, the client created on the first call will be returned for all subsequent calls. This is intentional for connection pooling - each provider should use consistent settings. The ProviderRegistry ensures consistent overrides per provider type. Args: name: Provider name (e.g., ‘voyage’, ‘cohere’, ‘qdrant’). **overrides: Override default limits/timeouts for this client: - max_connections: int - max_keepalive_connections: int - keepalive_expiry: float - connect_timeout: float - read_timeout: float - write_timeout: float - pool_timeout: float - http2: bool (default True) Returns: Configured httpx.AsyncClient with connection pooling. ### Method: `get_client_sync` [Section titled “Method: get_client_sync”](#method-get_client_sync) ** ``` get_client_sync() ``` Synchronous version of get_client for non-async contexts. This method is thread-safe via double-checked locking pattern. It can be called concurrently from multiple threads during provider initialization without creating duplicate clients. Note: Clients are cached by name only, not by override parameters. If the same provider name is requested multiple times with different overrides, the client created on the first call will be returned for all subsequent calls. This is intentional for connection pooling - each provider should use consistent settings. Warning: Synchronous and asynchronous methods should NOT be mixed. The locking strategy does not protect against concurrent access from both threads and coroutines. Only use sync methods in a purely threaded context, and async methods in a purely async context. Args: name: Provider name (e.g., ‘voyage’, ‘cohere’, ‘qdrant’). **overrides: Override default limits/timeouts for this client. Returns: Configured httpx.AsyncClient with connection pooling. ### Method: `get_instance` [Section titled “Method: get_instance”](#method-get_instance) ** ``` get_instance() ``` Get or create the singleton HttpClientPool instance. This method is thread-safe via double-checked locking pattern. Args: limits: Optional connection pool limits (only used on first call). timeouts: Optional timeout configuration (only used on first call). Returns: The singleton HttpClientPool instance. ### Method: `has_client` [Section titled “Method: has_client”](#method-has_client) ** ``` has_client() ``` Check if a client exists for the given provider name. Args: name: Provider name to check. Returns: True if a client exists for this provider. ### Method: `reset_instance` [Section titled “Method: reset_instance”](#method-reset_instance) ** ``` reset_instance() ``` Reset the singleton instance (primarily for testing). ## Class: `PoolLimits` [Section titled “Class: PoolLimits”](#class-poollimits) HTTP connection pool limits configuration. Attributes: max_connections: Maximum total connections across all hosts. max_keepalive_connections: Maximum persistent connections to keep alive. keepalive_expiry: Seconds to keep idle connections alive. ## Class: `PoolTimeouts` [Section titled “Class: PoolTimeouts”](#class-pooltimeouts) HTTP timeout configuration for pooled clients. Attributes: connect: Connection establishment timeout in seconds. read: Read timeout in seconds (longer for embedding/vector operations). write: Write timeout in seconds. pool: Pool acquire timeout in seconds. ## Function: `get_http_pool` [Section titled “Function: get_http_pool”](#function-get_http_pool) ** ``` get_http_pool() ``` Get the global HTTP client pool instance. This is the recommended way to access the HTTP client pool throughout the application. It returns the singleton instance, creating it if needed. Returns: The singleton HttpClientPool instance. ## Function: `reset_http_pool` [Section titled “Function: reset_http_pool”](#function-reset_http_pool) ** ``` reset_http_pool() ``` Reset the global HTTP client pool (primarily for testing). This async function properly closes all existing clients before resetting the singleton instance to prevent resource leaks. ## Function: `reset_http_pool_sync` [Section titled “Function: reset_http_pool_sync”](#function-reset_http_pool_sync) ** ``` reset_http_pool_sync() ``` Synchronous reset for testing fixtures (no cleanup). Warning: This does NOT close existing clients. Use reset_http_pool() if you need to clean up resources properly.