diff --git a/README.md b/README.md index f8fe4529..5b697352 100644 --- a/README.md +++ b/README.md @@ -270,6 +270,7 @@ Want to build your own skill on top of Fleet? See [docs/writing-skills.md](docs/ | Topic | Link | |-------|------| +| Codebase wiki (architecture, internals, AI Q&A) | [DeepWiki](https://deepwiki.com/Apra-Labs/apra-fleet) | | Install, uninstall, the `--llm` flag | [docs/install.md](docs/install.md) | | Choosing a provider | [docs/provider-guide.md](docs/provider-guide.md) | | FAQ | [docs/FAQ.md](docs/FAQ.md) | diff --git a/ROADMAP.md b/ROADMAP.md index 3d3e3b68..45faa167 100644 --- a/ROADMAP.md +++ b/ROADMAP.md @@ -3,38 +3,39 @@ This roadmap reflects current plans. Priorities shift based on community feedback. Have an idea? [Open a feature request](https://github.com/Apra-Labs/apra-fleet/issues/new/choose). -Items marked 🌱 are good first issues for new contributors. +Items marked [*] are good first issues for new contributors. --- ## Near-term (next 1-2 sprints) -- [ ] **npm publish** — Publish to npm so users can `npx apra-fleet install` 🌱 -- [ ] **Cursor / Windsurf integration guide** — Step-by-step docs for non-Claude-Code AI assistants 🌱 -- [ ] **Codex CLI & GitHub Copilot support** — First-class support as member agent providers — coming soon 🚀 -- [ ] **Member groups / tags** — Tag members (e.g. `gpu`, `build`, `test`) and target prompts at groups -- [ ] **Session log export** — Export prompt execution history as markdown 🌱 +- [ ] **npm publish** -- Publish to npm so users can `npx apra-fleet install` [*] +- [ ] **HTTP+SSE transport** -- MCP HTTP+SSE transport enabling remote members to connect over the network +- [ ] **Web dashboard** -- Browser UI for fleet topology, member status, live prompt monitoring, token/cloud cost visibility, git activity, and audit trail; VS Code extension embeds the same view with clickable file paths ([discussion #188](https://github.com/Apra-Labs/apra-fleet/discussions/188)) +- [ ] **Memory plane integration** -- Connect fleet agents to codebase knowledge indices so task planning starts from a structured map of architecture and relevant files rather than blind token-expensive discovery; encompasses persistent inter-session memory and sprint-accumulated knowledge graphs ([RFC #265](https://github.com/Apra-Labs/apra-fleet/discussions/265), [discussion #249](https://github.com/Apra-Labs/apra-fleet/discussions/249), [PR #266](https://github.com/Apra-Labs/apra-fleet/pull/266)) +- [ ] **Session log export** -- One-click markdown export of `fleet logs` output from the dashboard; builds on session history and listing work ([issue #189](https://github.com/Apra-Labs/apra-fleet/issues/189)) [*] +- [ ] **Member groups / tags** -- Tag members (e.g. `gpu`, `build`, `test`) and target prompts at groups; categorized grouping in status output ([PR #238](https://github.com/Apra-Labs/apra-fleet/pull/238)) -## Medium-term (3-6 months) +## Medium-term (1-2 months) -- [ ] **Expanded member targets** — Docker containers, Kubernetes pods, WSL instances, Azure VMs, and GCP VMs as first-class fleet members -- [ ] **Web dashboard** — Browser UI for fleet status and member management (foundation in `dashboard/`) 🌱 -- [ ] **Plugin system** — Community-built plugins for custom tools and providers -- [ ] **Result aggregation** — Merge outputs from parallel member executions into a unified report -- [ ] **Cost dashboard** — Visualize token and cloud spend across members +- [ ] **Agent-to-agent communication** -- Members communicate directly, not only through coordinator ([discussion #196](https://github.com/Apra-Labs/apra-fleet/discussions/196), [issue #152](https://github.com/Apra-Labs/apra-fleet/issues/152)) +- [ ] **Playbooks** -- JIT-compiled orchestration sequences: write environment-neutral runbooks once, fleet compiles them to native scripts per device and caches them; zero LLM cost on repeat runs with autonomous repair on failure ([discussion #194](https://github.com/Apra-Labs/apra-fleet/discussions/194)) +- [ ] **PM as full product lifecycle manager** -- Extend the PM skill through deploy, integration verification, quality grading, log monitoring, and backlog reprioritization; closes the loop beyond PR approval ([discussion #198](https://github.com/Apra-Labs/apra-fleet/discussions/198)) +- [ ] **Extension layer** -- Org-private skills and template overrides that survive fleet updates via a shadowing directory (`~/.apra-fleet/extensions/`); foundation for a community marketplace ([discussion #195](https://github.com/Apra-Labs/apra-fleet/discussions/195)) +- [ ] **Expanded member targets** -- Docker containers, Kubernetes pods, WSL instances, Azure VMs, and GCP VMs as first-class fleet members; includes per-instance data dir isolation for multi-fleet on the same machine ([PR #231](https://github.com/Apra-Labs/apra-fleet/pull/231)) +- [ ] **Multiple providers per member** -- Assign more than one LLM provider to a member and switch between them without re-registration ([issue #125](https://github.com/Apra-Labs/apra-fleet/issues/125)) +- [ ] **Slack notifications** -- Notify a Slack channel when members go offline, get blocked, or complete a task ([issue #77](https://github.com/Apra-Labs/apra-fleet/issues/77)) -## Long-term (6+ months) +## Long-term (3+ months) -- [ ] **Agent-to-agent communication** — Members communicate directly, not only through coordinator -- [ ] **Distributed task queue** — Priority-based job scheduling with retry and dead-letter handling -- [ ] **Fleet-as-a-Service** — Hosted version for teams that don't want to self-host -- [ ] **Terraform / Pulumi provider** — Infrastructure-as-code for fleet topology -- [ ] **Audit log** — Immutable log of all fleet operations for enterprise governance 🌱 -- [ ] **Multi-fleet federation** — Connect multiple fleet servers for cross-team orchestration +- [ ] **Fleet-as-a-Service** -- Hosted version for teams that don't want to self-host +- [ ] **Multi-fleet federation** -- Connect multiple fleet servers for cross-team orchestration +- [ ] **Terraform / Pulumi provider** -- Infrastructure-as-code for fleet topology +- [ ] **Audit log** -- Immutable append-only record of all fleet operations and secret usage for enterprise governance ([issue #159](https://github.com/Apra-Labs/apra-fleet/issues/159)) [*] --- ## Contributing -Pick any item above (especially 🌱 ones), open an issue to discuss your approach, then submit a PR. +Pick any item above (especially [*] ones), open an issue to discuss your approach, then submit a PR. See [CONTRIBUTING.md](CONTRIBUTING.md) for guidelines. diff --git a/docs/FAQ.md b/docs/FAQ.md index 9575f93e..bbf9dfa7 100644 --- a/docs/FAQ.md +++ b/docs/FAQ.md @@ -10,6 +10,19 @@ All questions and answers are maintained at: **[FAQ Index -- GitHub Discussions #127](https://github.com/Apra-Labs/apra-fleet/discussions/127)** +--- + +Want to explore the codebase or ask questions about how Fleet works internally? +The **[Fleet DeepWiki](https://deepwiki.com/Apra-Labs/apra-fleet)** has an AI assistant trained on the Fleet source code. + +Some useful starting points: + +- [How does execute_prompt work end-to-end?](https://deepwiki.com/Apra-Labs/apra-fleet/4.2-work-execution-tools) +- [How do I register and manage fleet members?](https://deepwiki.com/Apra-Labs/apra-fleet/4.1-member-lifecycle-tools) +- [What is the doer-reviewer workflow?](https://deepwiki.com/Apra-Labs/apra-fleet/7.1-sprint-lifecycle-and-doer-reviewer-loop) + +--- + Topics covered: - **Getting started** -- installation, device requirements, provider support diff --git a/docs/site/icons/icon.svg b/docs/site/icons/icon.svg deleted file mode 100644 index 1817854b..00000000 --- a/docs/site/icons/icon.svg +++ /dev/null @@ -1,36 +0,0 @@ - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - \ No newline at end of file diff --git a/docs/site/index.html b/docs/site/index.html deleted file mode 100644 index fac479d3..00000000 --- a/docs/site/index.html +++ /dev/null @@ -1,790 +0,0 @@ - - - - - `n - Apra Fleet — Your AI Engineering Organization - - - - - -
-
- - -
-
- - -
-
-
-

Your AI Engineering Organization

-

Transforms AI coding agents into a managed engineering team — with a PM that plans, doers that build, and reviewers that verify. On your machines, with your process.

-
- Get Started - View on GitHub -
-
-
- 3 - Agent Roles -
-
- 7 - Fleet Members -
-
- 100% - Your Infrastructure -
-
-
-
-
- -
-
┌─ claude ─────────── project: apra-fleet ─┐
-
-
> /pm status fleet
-
-
🔵 fleet-dev1 (idle)  · local  · windows
-
🟢 fleet-dev2 (idle)  · 192.168.1.13  · macos
-
🟢 fleet-rev1 (idle)  · 192.168.1.13  · macos
-
Fleet v0.1.2  ·  7/7 online
-
-
-
- -
-
┌─ claude ─────────── project: apra-fleet ─┐
-
-
> /pm plan #67
-
-
Planning Sprint 6a…
-
✓ Requirements written
-
✓ Plan generated (7 tasks, 3 phases)
-
✓ Reviewer APPROVED (12/12 checks)
-
Ready for execution.
-
-
-
- -
-
✨ Gemini ─────────── project: apra-fleet ─┐
-
-
> /pm start dev1 plan
-
-
🔵 fleet-dev1: Executing Phase 1…
-
✓ Task 1: Schema + data layer — committed
-
✓ Task 2: UI implementation — committed
-
⏸ VERIFY-1: Tests passing (524/524)
-
Stopped at checkpoint. Awaiting review.
-
-
-
- -
-
✨ Gemini ─────────── project: apra-fleet ─┐
-
-
> /pm deploy dev1
-
-
Deploying to production…
-
✓ Health check: ok
-
✓ Version: 0.2.4
-
✓ Deployed successfully
-
Production: live 🟢
-
-
-
-
-
-
-
- - -
-
-
HOW IT WORKS
-

The Fleet Architecture

-

- Three specialized roles. One unified workflow. Everything through git. -

-
-
-
📋
-

PM (Project Manager)

-

Plans sprints, breaks issues into tasks, assigns work, tracks progress. Uses a premium model for deep reasoning.

-
/pm plan, /pm start, /pm status
-
-
-
🔨
-

Doer (Specialist)

-

Executes implementation plans step by step. Roles include architect, designer, developer, integration test writer, DevOps engineer, and documentation writer.

-
git commit, test, verify
-
-
-
🔍
-

Reviewer (Quality Gate)

-

Reviews every PR against the plan. Blocks merges that miss requirements. Thorough analysis at every checkpoint.

-
review, approve, block
-
-
-
-
- - -
-
-

Why Apra Fleet?

-
-
-
🌐
-

Cross-Platform Fleet

-

Windows, macOS, Linux. Local members + remote via SSH. Mix and match your infrastructure.

-
-
-
🧠
-

Smart Model Routing

-

Route premium models for planning, mid-tier for coding, fast models for status checks. The right tier for each task.

-
-
-
🔀
-

Git as Transport

-

Everything flows through git. No proprietary sync protocol. Standard branches and PRs.

-
-
-
🛡️
-

Crash Recovery

-

/pm recover inspects git state and resumes from last checkpoint. Nothing is lost.

-
-
-
☁️
-

Cloud Compute

-

AWS EC2 auto start/stop. GPU-aware idle detection. Pay only when building.

-
-
-
✅
-

Mandatory Review Gates

-

No reviewer approval = no merge. Every change gets verified against the plan.

-
-
-
-
- - -
-
-

Get Running in 3 Steps

-
-
-
1
-

Download

-

Available for Windows, Ubuntu, and macOS. Download the latest release from GitHub.

-
-
-
2
-

Install & Register

-

Install the executable, then register fleet members — local machines or remote via SSH.

-
-
-
3
-

Start Managing

-

/pm plan to generate a plan, /pm start to build, /pm deploy to ship.

-
-
-
- Download Latest Release -
-
-
- - -
-
-

How Fleet Compares

-
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
CategoryToolsWhat’s MissingApra Fleet
Better AutocompleteCursor, Copilot, Windsurf, ClineOrganization, process, scaleFull engineering org
Autonomous Task RunnerDevin, Codex, OpenHandsYour infrastructure, your processRuns on your machines
Multi-Agent FrameworkCrewAI, AutoGen, LangGraph, MetaGPTWorking system, git workflowsProduction-ready system
AI-Native Engineering OrgApra Fleet—The complete picture
-
-
-
- - -
-
-

Real Stories

-
-
-

The Weekend Rewrite

-

CTO demands auth system migration by Monday. Fleet plans, builds, and ships the rewrite overnight. 3 doers, 2 reviewers, zero humans coding.

-
-
-

The Solo Founder

-

200-item backlog, one person. Ships 5 features per week writing zero code. PM plans each sprint, doers execute, reviewers verify.

-
-
-

The 3am Recovery

-

Production PM crashes mid-sprint. /pm recover inspects git state, finds last checkpoint, resumes all in-flight work. Nothing lost.

-
-
-
-
- - -
-
-

Running in Production

- -

Battle-tested across C++, Node.js, Python, ML Pipelines, and Video Processing projects

-
-
- - - - -