v0.3.0

What's New

Slack Integration (--features slack)

Full Slack bot via Socket Mode — slash commands (/sparks help, status, plan, implement, review, explain, search, alerts, ghosts, memories, dispatch, model, jobs, mood, knobs, session, cli, set, watch), Block Kit planning interview, streaming responses, per-channel auth and rate limiting, confirmations with approve/deny buttons, and pulse delivery to channels.

Microsoft Teams Integration (--features teams)

Bot Framework REST API with JWT RS256 signature verification, serviceUrl validation against known Bot Framework domains, tenant authorization (allowed_tenants / allow_all_tenants), Adaptive Cards for confirmations and planning interview, bearer token cache with 60s pre-expiry window, and all equivalent commands.

Proactive Alerting Engine

Rule-based alert evaluation against the activity log with multi-channel delivery (log, slack, teams, webhook), configurable severity thresholds (info / warning / critical), and silence windows to suppress repeated alerts. Configure via [alerts] in config.toml.

Semantic Memory Deduplication and Decay Scoring

Exponential decay scoring with configurable half-life (decay_half_life_days), soft entry cap (max_entries), and a deduplication pass to prune near-duplicate memories — keeps memory search relevant over time.

SonarQube MCP Quality Gate Integration

Poll SonarCloud or self-hosted SonarQube quality gates from within Sparks tasks. Optionally blocks on gate failure. Configure via [sonarqube] in config.toml or SPARKS_SONAR_TOKEN env var.

Workspace Snapshot and Time-Travel Debugging

sparks snapshot create|list|diff|restore — point-in-time workspace snapshots with configurable retention limits, size guard, include/exclude globs, and atomic restore. Configure via [snapshot] in config.toml.

Ghost Performance Leaderboard and A/B Testing

sparks leaderboard show|compare — track ghost success rates, run A/B routing between ghosts with a configurable traffic split, and surface promotion recommendations when a challenger outperforms the incumbent. Configure via [leaderboard] in config.toml.

Other Changes

• SPARKS_SLACK_BOT_TOKEN, SPARKS_SLACK_APP_TOKEN, SPARKS_SLACK_SIGNING_SECRET registered in the secrets registry — use sparks secrets set slack.* or the OS keyring.
• Complete Athena→Sparks rebrand across env vars, binary references, config paths (~/.sparks/), CLI contract tag ([sparks_cli_contract]), and all docs.

Setup Guides

• Slack setup
• Teams setup

v0.2.0

Full Changelog: v0.1.2...v0.2.0

v0.1.2

Full Changelog: v0.1.0...v0.1.2

v0.1.0 — Initial public release

Athena v0.1.0 — Initial Public Release

First public release of Athena, a self-hosted multi-agent orchestrator built in Rust.

What's included

Core runtime

• Multi-ghost agent architecture — configurable specialized agents (coder, scout, self-dev) with role-based tool access
• Hardened Docker execution sandbox — CAP_DROP ALL, read-only rootfs, no network by default
• OpenAI-compatible API endpoint — drop-in for tools that speak the OpenAI chat completions protocol
• Three execution strategies: react, code (CLI-driven), and precheck

Memory & context

• HNSW semantic memory index with exact fallback for fast nearest-neighbour recall
• Adaptive context budgeting — dynamically fits conversation history into the token window
• LRU hot-path cache on the memory read path

Safety & observability

• Input-layer prompt scanner with flag_only / block modes, allowlist, and per-provider overrides
• SHA-256 tool-call loop guard — circuit breaker against infinite tool loops
• Full observer event bus with 20 categorised event types
• Session review & explainability — /review, /explain, /watch, /alerts via Telegram

Self-improvement

• KPI tracking — task success rate, verification pass rate, rollback rate, MTTF
• Eval harness with versioned scenario suites and scorecard diffs
• Optimizer tournament — nightly KPI-driven ghost selection
• Feature contract DAG — multi-task dependency ordering with rollback

Integrations

• MCP (Model Context Protocol) tool registry — stdio transport, namespaced tools, allowlist
• Ticket intake — Linear, GitHub, GitLab, Jira, webhooks
• Langfuse tracing, CI monitor, scheduled jobs

Quick start

git clone https://github.com/Enreign/athena
cd athena
cp config.example.toml config.toml
# edit config.toml — add your LLM API key
cargo run -- ghosts        # verify agent roster
cargo run -- doctor        # end-to-end health check
cargo run -- dispatch --goal "your task here"

See the documentation for full setup and configuration.

Notes

• Requires Rust stable and Docker
• Telegram integration requires --features telegram
• This is an early release — internals will change, APIs are not yet stable