Skip to content
Selu Docs

Roadmap

This page lists features we’re planning or considering for future releases. Nothing here is guaranteed — priorities can shift based on feedback and technical constraints.

Have thoughts on what should come next? Join the discussion on GitHub or Discord.

Pipes

Section titled “Pipes”
  • Test connection button — One-click connection test during pipe setup for Telegram, iMessage, and other external services.
  • Telegram polling mode — Alternative to webhooks for Selu instances that aren’t reachable from the internet.
  • Telegram bot commands — Support for /start, /reset, /help and custom commands registered via BotFather.
  • Telegram group chat support — Allow a single Telegram pipe to serve multiple group chats, not just a single chat.
  • iMessage media support — Receive and process images and files sent through iMessage.
  • Approval timeouts — Automatically expire pending tool approval requests after a configurable time period.

Web Chat

Section titled “Web Chat”
  • Display name and welcome message — Customize the chat window title and the first message new users see.
  • Theme settings — Choose between light, dark, or system-matched appearance for the web chat.
  • Session timeout — Automatically reset agent context after a period of inactivity.
  • Message limits and rate limiting — Restrict message length and sending frequency.
  • Security settings page — Dedicated UI for managing authentication and access control.

Agent Routing

Section titled “Agent Routing”
  • Keyword-based routing — Route messages containing specific words to a specialized agent automatically.
  • Prefix command routing — Direct messages to agents using /agent-name syntax.
  • Per-pipe agent assignment — Assign different default agents to different pipes via a settings UI.
  • Agent selector in web chat — Pick an agent from a dropdown in the chat interface.

Triggers

Section titled “Triggers”
  • Time-based triggers — Activate a specific agent at certain times of day (e.g., a “morning briefing” agent at 8 AM).
  • Event-driven agent switching — Automatically switch to a specialized agent when a particular event occurs.
  • Handoff triggers — Let an agent hand a conversation off to another agent permanently when it detects a topic outside its expertise.

Event System

Section titled “Event System”
  • User-configurable event subscriptions — Subscribe to event types and choose which pipe receives notifications, with optional agent processing.
  • Event filtering — Filter events by severity, source, or content pattern before they reach you.
  • Event history viewer — Browse and search past events for debugging and auditing.

LLM Providers

Section titled “LLM Providers”
  • Ollama integration — Support for Ollama as a local model provider, so you can point Selu at a local Ollama instance without needing an API key.
  • Anthropic Top-p setting — Expose a Top-p parameter for fine-tuning Anthropic response randomness as an alternative to temperature.
  • OpenAI Organization ID — Allow specifying an Organization ID for OpenAI API keys that belong to multiple organizations.
  • Bedrock IAM credential authentication — Support AWS Access Key ID and Secret Access Key pairs for native IAM authentication, as an alternative to the current bearer token approach.

Database

Section titled “Database”
  • PostgreSQL support — Option to use PostgreSQL instead of the embedded SQLite database, for deployments that need a dedicated database server.

Agent Runtime

Section titled “Agent Runtime”
  • Agent container network configuration — Configure the Docker network name for agent containers.
  • Agent call timeout — Set a maximum time for agent responses before the orchestrator cancels the request.
  • Agent memory limits — Set memory limits per agent container.

Sessions

Section titled “Sessions”
  • Session idle timeout — Automatically reset agent context after a configurable period of inactivity.
  • Session history limits — Cap the number of messages kept in session context.

Environment Variable Configuration

Section titled “Environment Variable Configuration”
  • LLM provider credentials via environment variables — Configure Anthropic, OpenAI, and Bedrock API keys through environment variables as an alternative to the dashboard UI.
  • Channel credentials via environment variables — Configure Telegram bot tokens and BlueBubbles/iMessage settings through environment variables as an alternative to the dashboard UI.

Conversation Inspector

Section titled “Conversation Inspector”
  • Conversation inspector — A built-in panel that shows what happens behind the scenes during a conversation: the messages exchanged, capability calls, and raw model responses. Useful for understanding agent behavior and debugging custom agents.

Security

Section titled “Security”
  • Audit log — A log of tool usage, approval decisions, and security-relevant events. Useful for understanding what your agents are doing and reviewing activity over time.
  • Email-based password recovery — Allow users to reset their own password via email instead of requiring an administrator to use the command line.

Settings UI

Section titled “Settings UI”
  • Routing configuration UI — Configure agent routing rules from a settings page.
  • Session settings UI — Manage idle timeouts and session behavior from the web interface.
  • Trigger configuration UI — Set up and manage triggers from a settings page.

Developer Tooling (CLI)

Section titled “Developer Tooling (CLI)”
  • selu init — Scaffold a new agent project with sensible defaults.
  • selu validate — Validate agent.yaml and manifest.yaml against their schemas before publishing.
  • selu dev — Local development server that loads your agent and opens a web chat for testing, with --watch for auto-reload on file changes.
  • selu chat — Send a single message to an agent from the command line for quick, non-interactive testing.
  • selu logs — Stream orchestrator and capability logs with filtering by capability name and log level.
  • selu status — Show the state of running capability containers.
  • selu test snapshot — Record and replay conversations for regression testing.
  • selu publish — Publish an agent package to the Selu marketplace.
  • selu build — Build capability Docker images locally.

Template Variables in System Prompts

Section titled “Template Variables in System Prompts”
  • Runtime template substitution in agent.md — Support variables like {{user_name}}, {{agent_name}}, {{date}}, and {{capabilities}} that are resolved at runtime before the system prompt is sent to the LLM.

Advanced Routing

Section titled “Advanced Routing”
  • Intent-based routing — Lightweight classifier that detects semantic intent (e.g., weather_query) to route messages to the correct agent.
  • Regex-based routing — Match messages against regular expressions for precise routing control.
  • Routing priority scoring — When multiple agents match, use a configurable priority value to break ties.

Event System Enhancements

Section titled “Event System Enhancements”
  • emit_event as an agent tool — Wire the existing emit_event implementation into the agent tool list so agents can emit events during conversations. The backend code exists but is not yet connected.
  • Circular delegation detection — Detect and block A → B → A delegation loops to prevent infinite recursion.
  • Delegation depth and rate limits — Configurable maximum delegation depth and per-session delegation limits.
  • Event rate limiting — Limits on events per agent per hour and maximum pending events.

Marketplace

Section titled “Marketplace”
  • Human review pipeline — Automated and manual review process for agent submissions before they appear in the marketplace.
  • Granular upgrade policies — Per-level update preferences (auto-patch, auto-minor, manual, pin) instead of a simple on/off toggle.
  • Version constraint syntax — SemVer constraint syntax (^, ~, >=) for capability image references in agent.yaml.
  • Pre-release version handling — Beta badge and opt-in for pre-release versions.

Webhook System

Section titled “Webhook System”
  • Outbound webhooks — Push events to external URLs when specific things happen in Selu (conversations, agent events, capability invocations).
  • Inbound webhook API — Structured inbound webhook endpoint with HMAC signature verification for external system integrations.
  • Webhook retry policies — Exponential backoff retry for failed webhook deliveries.

Public REST API

Section titled “Public REST API”
  • Bearer token authentication — API token-based authentication for programmatic access, as an alternative to session cookies.
  • Versioned API endpoints — Stable, versioned API at /api/v1/ for managing agents, conversations, and capabilities.
  • OpenAPI specification — Auto-generated OpenAPI spec for API exploration and client generation.

Agent Configuration

Section titled “Agent Configuration”
  • Per-agent memory settings — Configure session TTL, max history messages, and long-term memory toggle per agent in agent.yaml.
  • Per-agent max_tokens — Configure maximum response tokens per agent.
  • Model fallback — Specify a fallback model when the primary model is unavailable.