Cameron Cordes
efd05db523
Trialing Qwen3-Embedding-0.6B (1024-dim, instruct-prefixed queries) against nomic required code changes at every hardcoded seam; now it's a config flip plus a reembed_embeddings run. - EMBEDDING_DIM env (default 768) replaces every hardcoded dim check: daily summary / calendar / search / location DAOs, Ollama batch validation, reembed_embeddings - entities gains the dim guard it never had — a wrong-dim vector silently kills dedup/recall (cosine over mismatched lengths is 0), so store None and warn instead - embed_query / embed_document split with EMBED_QUERY_PREFIX / EMBED_DOCUMENT_PREFIX (literal \n expanded): retrieval models treat the two sides differently — nomic wants search_query:/search_document:, Qwen3 wants Instruct:...\nQuery: on queries only. All query-side call sites and all corpus writers now declare their side. - document the contract in CLAUDE.md: change the model or any of these vars → re-run reembed_embeddings or search is garbage Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
89 lines
3.4 KiB
Rust
89 lines
3.4 KiB
Rust
//! Bundle of the local LLM pair (Ollama + optional llama-swap) with the
|
|
//! `LLM_BACKEND` dispatch baked in.
|
|
//!
|
|
//! Exists because passing the pair around as loose values invited the same
|
|
//! bug three times: import/backfill tooling embedded corpora via
|
|
//! `OllamaClient` directly while the query side dispatched through
|
|
//! `embed_one`, so flipping `LLM_BACKEND=llamacpp` silently split queries
|
|
//! and corpus into different vector spaces. Anything that writes or reads
|
|
//! embeddings should go through this type (or `embed_one`/`embed_many`),
|
|
//! never a concrete client.
|
|
//!
|
|
//! Deliberately knows nothing about chat policy — hybrid/OpenRouter routing
|
|
//! is request-scoped and stays in `ResolvedBackend`. This is only the
|
|
//! local stack: embeddings and offline single-shot generation.
|
|
|
|
// Constructed by binaries, not the server — dead code from main.rs's view.
|
|
#![allow(dead_code)]
|
|
|
|
use std::sync::Arc;
|
|
|
|
use anyhow::Result;
|
|
|
|
use super::llamacpp::LlamaCppClient;
|
|
use super::llm_client::LlmClient;
|
|
use super::ollama::{EMBEDDING_MODEL, OllamaClient};
|
|
|
|
#[derive(Clone)]
|
|
pub struct LocalLlm {
|
|
ollama: OllamaClient,
|
|
llamacpp: Option<Arc<LlamaCppClient>>,
|
|
}
|
|
|
|
impl LocalLlm {
|
|
pub fn new(ollama: OllamaClient, llamacpp: Option<Arc<LlamaCppClient>>) -> Self {
|
|
Self { ollama, llamacpp }
|
|
}
|
|
|
|
/// Construct from the canonical env wiring shared with `AppState`.
|
|
pub fn from_env() -> Self {
|
|
Self::new(
|
|
crate::state::build_ollama_from_env(),
|
|
crate::state::build_llamacpp_from_env(),
|
|
)
|
|
}
|
|
|
|
/// Embed a search query (applies `EMBED_QUERY_PREFIX`). Callers must
|
|
/// pick query vs document — retrieval models treat the two sides
|
|
/// differently and an unmarked embed invites prefix-mismatch bugs.
|
|
pub async fn embed_query(&self, text: &str) -> Result<Vec<f32>> {
|
|
super::embed_query(&self.ollama, self.llamacpp.as_deref(), text).await
|
|
}
|
|
|
|
/// Embed corpus text (applies `EMBED_DOCUMENT_PREFIX`).
|
|
pub async fn embed_document(&self, text: &str) -> Result<Vec<f32>> {
|
|
super::embed_document(&self.ollama, self.llamacpp.as_deref(), text).await
|
|
}
|
|
|
|
/// Single-shot local text generation via the `LLM_BACKEND`-selected
|
|
/// client (offline tooling; chat turns belong to `ResolvedBackend`).
|
|
pub async fn generate(&self, prompt: &str, system: Option<&str>) -> Result<String> {
|
|
if super::local_backend_is_llamacpp() {
|
|
if let Some(lc) = self.llamacpp.as_deref() {
|
|
return <LlamaCppClient as LlmClient>::generate(lc, prompt, system, None).await;
|
|
}
|
|
anyhow::bail!(
|
|
"LLM_BACKEND=llamacpp but LlamaCppClient is unconfigured — \
|
|
set LLAMA_SWAP_URL or switch to LLM_BACKEND=ollama"
|
|
);
|
|
}
|
|
self.ollama.generate(prompt, system).await
|
|
}
|
|
|
|
/// Label identifying which backend + model produces embeddings right
|
|
/// now. Store it alongside vectors (`model_version` columns) so a
|
|
/// backend flip is detectable in the data, not just in env history.
|
|
pub fn embedding_model_version(&self) -> String {
|
|
if super::local_backend_is_llamacpp() {
|
|
let slot = self
|
|
.llamacpp
|
|
.as_deref()
|
|
.map(|c| c.embedding_model.as_str())
|
|
.unwrap_or("embed");
|
|
format!("llama-swap:{}", slot)
|
|
} else {
|
|
EMBEDDING_MODEL.to_string()
|
|
}
|
|
}
|
|
}
|