forge-guardrails 0.1.2

Foundation types for an LLM-agent workflow framework
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37

//! HTTP proxy interfaces, request handlers, and servers.

/// Low-level endpoint intercept handlers and validation logic.
pub mod handler;
/// Anthropic/OpenAI protocol translations and payload shaping.
#[allow(clippy::module_inception)]
pub mod proxy;
mod response;
/// HTTP server lifecycle and endpoints.
pub mod server;

pub use handler::{
    handle_anthropic_messages, handle_anthropic_messages_with_scorer,
    handle_anthropic_messages_with_scorers,
    handle_anthropic_messages_with_scorers_and_tool_controls,
    handle_anthropic_messages_with_scorers_and_tool_output_compression,
    handle_anthropic_messages_with_scorers_tool_controls_and_headers, handle_chat_completions,
    handle_chat_completions_with_scorer, handle_chat_completions_with_scorers,
    handle_chat_completions_with_scorers_and_tool_controls,
    handle_chat_completions_with_scorers_and_tool_output_compression,
    init_proxy_classifier_log_sink_from_env, init_proxy_tool_output_compression_log_sink_from_env,
    init_proxy_training_capture_sink_from_env, shutdown_proxy_classifier_log_sink,
    shutdown_proxy_tool_output_compression_log_sink, shutdown_proxy_training_capture_sink,
    AnthropicEventStream, AnthropicHandlerError, AnthropicHandlerResult, HandlerError,
    HandlerResult, OpenAiEventStream,
};
pub use proxy::{
    extract_passthrough, extract_sampling, has_respond_tool, openai_to_messages,
    respond_tool_openai, strip_respond_calls, text_response_to_openai,
    text_response_to_openai_with_usage, text_to_sse_events, text_to_sse_events_with_usage,
    tool_calls_to_openai, tool_calls_to_openai_with_usage, tool_calls_to_sse_events,
    tool_calls_to_sse_events_with_usage, OpenAiMessageError,
};
pub(crate) use proxy::{
    text_response_to_openai_with_usage_details, tool_calls_to_openai_with_usage_details,
};
pub use server::HTTPServer;