103 lines
3.2 KiB
Rust
103 lines
3.2 KiB
Rust
|
// Local Development Setup with Ollama
|
||
|
|
// No API keys required - uses local Ollama for LLM and embeddings
|
||
|
|
|
||
|
|
use std::sync::Arc;
|
||
|
|
|
||
|
|
use surrealdb::engine::remote::ws::Ws;
|
||
|
|
use surrealdb::opt::auth::Root;
|
||
|
|
use surrealdb::Surreal;
|
||
|
|
use vapora_llm_router::providers::OllamaClient;
|
||
|
|
use vapora_rlm::chunking::{ChunkingConfig, ChunkingStrategy};
|
||
|
|
use vapora_rlm::embeddings::EmbeddingConfig;
|
||
|
|
use vapora_rlm::engine::RLMEngineConfig;
|
||
|
|
use vapora_rlm::search::bm25::BM25Index;
|
||
|
|
use vapora_rlm::storage::SurrealDBStorage;
|
||
|
|
use vapora_rlm::RLMEngine;
|
||
|
|
|
||
|
|
#[tokio::main]
|
||
|
|
async fn main() -> anyhow::Result<()> {
|
||
|
|
println!("🦙 Local RLM Setup with Ollama");
|
||
|
|
println!("Prerequisites:");
|
||
|
|
println!(" - SurrealDB: docker run -p 8000:8000 surrealdb/surrealdb:latest start");
|
||
|
|
println!(" - Ollama: brew install ollama && ollama serve");
|
||
|
|
println!(" - Model: ollama pull llama3.2\n");
|
||
|
|
|
||
|
|
// 1. Setup SurrealDB
|
||
|
|
let db = Surreal::new::<Ws>("127.0.0.1:8000").await?;
|
||
|
|
db.signin(Root {
|
||
|
|
username: "root",
|
||
|
|
password: "root",
|
||
|
|
})
|
||
|
|
.await?;
|
||
|
|
db.use_ns("local").use_db("rlm").await?;
|
||
|
|
|
||
|
|
// 2. Setup Ollama client (local, no API key needed)
|
||
|
|
let llm_client = Arc::new(OllamaClient::new(
|
||
|
|
"http://localhost:11434".to_string(),
|
||
|
|
"llama3.2".to_string(),
|
||
|
|
4096, // max_tokens
|
||
|
|
0.7, // temperature
|
||
|
|
)?);
|
||
|
|
|
||
|
|
// 3. Create storage and BM25 index
|
||
|
|
let storage = Arc::new(SurrealDBStorage::new(db));
|
||
|
|
let bm25_index = Arc::new(BM25Index::new()?);
|
||
|
|
|
||
|
|
// 4. Configure RLM engine for local development
|
||
|
|
let rlm_config = RLMEngineConfig {
|
||
|
|
chunking: ChunkingConfig {
|
||
|
|
strategy: ChunkingStrategy::Fixed,
|
||
|
|
chunk_size: 500,
|
||
|
|
overlap: 100,
|
||
|
|
},
|
||
|
|
embedding: Some(EmbeddingConfig::ollama("llama3.2")),
|
||
|
|
auto_rebuild_bm25: true,
|
||
|
|
max_chunks_per_doc: 5_000,
|
||
|
|
};
|
||
|
|
|
||
|
|
// 5. Create RLM engine with Ollama client
|
||
|
|
let engine = RLMEngine::with_llm_client(storage, bm25_index, llm_client, Some(rlm_config))?;
|
||
|
|
|
||
|
|
println!("✓ RLM Engine configured with Ollama\n");
|
||
|
|
|
||
|
|
// 6. Example: Analyze Rust code
|
||
|
|
let doc_id = "rust-example";
|
||
|
|
let content = r#"
|
||
|
|
fn fibonacci(n: u32) -> u32 {
|
||
|
|
match n {
|
||
|
|
0 => 0,
|
||
|
|
1 => 1,
|
||
|
|
_ => fibonacci(n - 1) + fibonacci(n - 2),
|
||
|
|
}
|
||
|
|
}
|
||
|
|
|
||
|
|
// This recursive implementation has exponential time complexity.
|
||
|
|
// A better approach would use dynamic programming or iteration.
|
||
|
|
"#;
|
||
|
|
|
||
|
|
println!("📄 Loading Rust code...");
|
||
|
|
let chunk_count = engine.load_document(doc_id, content, None).await?;
|
||
|
|
println!("✓ Loaded {} chunks\n", chunk_count);
|
||
|
|
|
||
|
|
println!("🔍 Searching for 'complexity'...");
|
||
|
|
let results = engine.query(doc_id, "complexity", None, 3).await?;
|
||
|
|
println!("✓ Found {} results\n", results.len());
|
||
|
|
|
||
|
|
println!("🦙 Asking Ollama to explain the code...");
|
||
|
|
let response = engine
|
||
|
|
.dispatch_subtask(
|
||
|
|
doc_id,
|
||
|
|
"Explain this Rust code and suggest improvements",
|
||
|
|
None,
|
||
|
|
3,
|
||
|
|
)
|
||
|
|
.await?;
|
||
|
|
println!("✓ Ollama says:\n{}\n", response.text);
|
||
|
|
println!(
|
||
|
|
" (Used {} tokens)",
|
||
|
|
response.total_input_tokens + response.total_output_tokens
|
||
|
|
);
|
||
|
|
|
||
|
|
Ok(())
|
||
|
|
}
|