mindfork 0.21.0

A terminal AI chat written in Rust: local models via llama.cpp, or OpenAI, Anthropic, Gemini, Grok and OpenRouter in the cloud, with persistent memory, notes, RAG and tools.
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
38
39

//! Inference backends of the OpenAI family. Two protocols:
//! - **Chat Completions** ([`OpenAiClient`], `wire`): local/external `llama-server`
//!   (and OpenAI-compatible proxies/embeddings), where sampling is sent as-is; xAI
//!   and DeepSeek, which take the same body (DeepSeek with its own thinking switch,
//!   docs/history/deepseek-mode.md); and the OpenRouter gateway, in a dialect of
//!   its own (docs/research/openrouter-mode.md). The other clouds moved to their own
//!   protocols, see ADR 0004;
//! - **Responses API** ([`ResponsesClient`], `responses`): OpenAI cloud
//!   (`platform.openai.com/v1/responses`) with reasoning summaries, `reasoning.effort`,
//!   and `text.verbosity`. See ADR 0004, docs/research/openai-responses-client.md.
//!
//! Both implement [`EngineBackend`](super::contract::EngineBackend);
//! [`Embedder`](super::contract::Embedder) — only [`OpenAiClient`] (`/v1/embeddings`;
//! Responses has no embeddings). Gemini cloud uses the native
//! [`GeminiClient`](super::gemini::GeminiClient), not this client.

#[cfg(test)]
mod cache_probe_tests;
pub mod client;
#[cfg(test)]
mod deepseek_live_tests;
#[cfg(test)]
mod deepseek_tests;
#[cfg(test)]
mod gateway_live_tests;
#[cfg(test)]
mod gateway_tests;
pub mod responses;
mod wire;
#[cfg(test)]
mod xai_live_tests;
#[cfg(test)]
mod xai_tests;

pub(crate) use client::attributed;
pub use client::{KeyVerdict, OpenAiClient};
pub use responses::ResponsesClient;
pub(crate) use wire::ModelEnvelope;
pub use wire::tools_json;