honcho-ai

Rust SDK for Honcho — AI agent memory and social cognition infrastructure.

Status: Beta — full API parity with Python/TypeScript SDKs. Production-ready for async Rust.

What is Honcho?

Honcho is an AI agent memory platform that provides persistent, contextual memory and social cognition for LLM agents. It enables your agents to:

• Remember past interactions across sessions and users
• Understand social context — who knows what about whom
• Reason dialectically — query and synthesize knowledge across peers
• Evolve self-representation and peer models over time

This crate is a native Rust client for the Honcho API (v3), with 100% parity with the official Python and TypeScript SDKs.

Installation

Add to your Cargo.toml:

[dependencies]
honcho-ai = "0.1"
tokio = { version = "1", features = ["full"] }

Or via cargo:

cargo add honcho-ai

Quickstart

use honcho_ai::Honcho;

#[tokio::main]
async fn main() -> honcho_ai::error::Result<()> {
    // Connect to Honcho (local or cloud)
    let client = Honcho::from_params(
        Honcho::builder()
            .base_url("http://localhost:8000")
            .workspace_id("my-agent")
            .build(),
    )?;

    // Create a peer (agent or user identity)
    let peer = client.peer("alice", None, None).await?;

    // Create a session and add a message
    let session = client.session("chat-1", None, None, None).await?;
    session.add_messages(vec![
        peer.message("Hello! My name is Alice and I like hiking.")
            .build()?
    ]).await?;

    // Ask the dialectic layer — Honcho reasons about what it knows
    let reply = peer.chat("What does Alice like?").await?;
    if let Some(text) = reply {
        println!("Alice likes: {text}");
    }

    // Streaming chat for real-time response delivery
    use futures_util::StreamExt;
    let mut stream = peer.chat_stream("Tell me more about Alice")
        .send()
        .await?;
    while let Some(chunk) = stream.next().await {
        print!("{}", chunk?);
    }

    Ok(())
}

Key Concepts

Core API

Client (Honcho)

let client = Honcho::new("http://localhost:8000", "my-workspace")?;

// Workspace management
client.force_ensure().await?;          // ensure workspace exists
client.get_metadata().await?;          // fetch metadata
client.set_metadata(map).await?;       // set metadata
client.get_configuration().await?;     // typed config
client.delete_workspace("old").await?; // delete workspace

// Paginated listing
let peers_page = client.peers().await?;
let sessions_page = client.sessions().await?;
let workspaces_page = client.workspaces().await?;

// Search & dream
let results = client.search("topic", None, None).await?;
client.schedule_dream("alice", None, None).await?;

Peer

let peer = client.peer("alice", None, None).await?;

// Dialectic chat
let reply = peer.chat("What does Alice like?").await?;
let reply = peer.chat_with_options(&opts).await?;
let stream = peer.chat_stream("Hello").send().await?;

// Representation & context
let rep = peer.representation().await?;
let ctx = peer.context().await?;
let ctx = peer.context_with_target("bob").await?;

// Peer card (key facts)
let card = peer.get_card().await?;
peer.set_card(vec!["friendly".into(), "rust-dev".into()]).await?;

// Conclusions
peer.conclusions().list().send().await?;
peer.conclusions().create(["Likes hiking"]).await?;
peer.conclusions().delete(conclusion_id).await?;
let cross = peer.conclusions_of("bob"); // cross-peer conclusions

// Search & sessions
let results = peer.search("topic").await?;
let page = peer.sessions().await?;

// Metadata CRUD
peer.refresh().await?;
peer.get_metadata().await?;
peer.set_metadata(map).await?;
peer.update(patch).await?;            // patch update

Session

let session = client.session("chat-1", None, None, None).await?;

// Messages (batch up to 100 per call)
let msgs = session.add_messages(vec![msg]).await?;
let page = session.messages().await?;
let msg = session.get_message("msg-1").await?;
let msg = session.update_message("msg-1", metadata).await?;

// File upload — typed sources (Rust-only feature)
use honcho_ai::FileSource;
let msgs = session.upload_file(
    FileSource::bytes("doc.pdf", data, "application/pdf")
).peer("alice").send().await?;

let msgs = session.upload_file(
    FileSource::path("report.txt")
).peer("alice").send().await?;

// Peer management
session.add_peer("bob").await?;
session.set_peers(["alice", "bob"]).await?;
session.remove_peers(["bob"]).await?;

// Context & search
let ctx = session.context().await?;
let results = session.search("topic").await?;
let summaries = session.summaries().await?;

// Clone & delete
let cloned = session.clone_session().await?;
session.delete().await?;

Error Handling

use honcho_ai::error::HonchoError;

match client.peer("alice", None, None).await {
    Ok(peer) => { /* ... */ },
    Err(e @ HonchoError::NotFound { .. }) => {
        eprintln!("Peer not found: {}", e.message());
        // e.code() returns "not_found"
        // e.status_code() returns Some(404)
        // e.is_retryable() returns false
    },
    Err(e) if e.is_retryable() => {
        // Timeout, connection error, or 429/5xx — retry
        tokio::time::sleep(e.retry_after().unwrap_or(Duration::from_secs(1))).await;
    },
    Err(e) => return Err(e),
}

The SDK auto-retries transient errors (2 retries, exponential backoff, respects Retry-After). You don't need to implement your own retry loop for timeouts, connection errors, 429, or 5xx responses.

Features

Blocking API

Enable the blocking feature for synchronous usage:

honcho-ai = { version = "0.1", features = ["blocking"] }

use honcho_ai::blocking::Honcho;

let client = Honcho::new("http://localhost:8000", "my-workspace")?;
let peer = client.peer("alice", None, None)?;
let reply = peer.chat("Hello")?;

Note: The blocking API uses an internal tokio runtime. It will panic if you call it from within an existing async context — use the async API instead.

Rust-Only Features

These APIs have no equivalent in the Python/TypeScript SDKs:

Pagination

All list endpoints return a Page<T> with 1-based page numbers:

// Default: page 1, 50 items
let page = client.peers().await?;
println!("Page {} of {} ({} total)", page.page(), page.pages(), page.total());
for peer in page.items() {
    println!("  - {}", peer.id);
}

// Auto-fetch all pages
use futures_util::StreamExt;
let stream = page.into_stream();
tokio::pin!(stream);
while let Some(Ok(item)) = stream.next().await {
    // process each item lazily
}

// Manual pagination
let mut page = client.sessions_with_filters(filters, 1, 10, false).await?;
while page.has_next() {
    page = page.next_page().await?;
}

Environment Variables

Explicit builder arguments take precedence over environment variables.

Parity with Official SDKs

Requirements

• Rust 1.88.0 or later (edition 2024)
• Tokio async runtime (multi-thread)
• Honcho server (local or api.honcho.dev)

Links

• Repository
• API Documentation
• Honcho Server
• OpenAPI Spec
• Migration Guide
• Changelog

License

MIT — see LICENSE.md.

Built by Leszek. Not affiliated with Plastic Labs — community-maintained with the intent to contribute upstream.