# Agora: ta-amps ecosystem guide (Track 2: Push the Frontier)

> Built for **Track 2: Push the Frontier** (Developer Tooling, Agentic Infrastructure & Security), `ta-amps` pairs one shared **durable task queue** of **durable tasks (`tasks (things to do)`)** and **`task objects`** with 7-profile constructible stem-cell DAG execution, `< 2 ms` Herma context shimming, `maker != checker` proof-carrying verification, Meta open-weights fine-tuning (`Muse Glimmer 30B` QLoRA, `Llama 3.2 1B/3B` DoRA / Iterative-RPO / SpinQuant, `Muse Spark 1.3`, `Muse Wonder 1.0`, `Muse Code`, `nanoMuse`, `Meta SAM 3.1`), and official partner compute integration (**Baseten** `meta-models/muse-spark-open`, **Fireworks AI**, and **Modal**), with **[Eranos](https://eranos.ta-amps.workers.dev)** as the live public-good showcase surface. Agora is one of exactly two public surfaces: the single overview of architecture and setup, 1x3 product hub, interactive durable task and task-object explorer, use cases, connecting any agent and contributing. Eranos, the two-sided marketplace, is the other. Tessera and Herma are infrastructure; Home is not a surface, and its address sends people to Agora. Everything else is nouns and verbs (`task objects` and `tasks (things to do)`) in seed packs, agent kits included. Public architecture and dated status. Documentation only, with no application runtime or credentials.

- [Overview](README.md): Track 2 framing, the two surfaces, Meta AI & partner compute stack table, the infrastructure, the pack library and the repository map.
- [For judges](https://agora.ta-amps.workers.dev/judges/) and [Live status](https://agora.ta-amps.workers.dev/status/): five-minute walkthrough, live health probes, replayed E2E receipts, 5 hosted feeds, 1,000 rule-checked scenarios, and all six repositories.
- [Required checkpoint](docs/architecture-checkpoint.md): nouns and verbs, two surfaces and narrow infrastructure; packs are unnamed bundles, never surfaces, and projects only labels.
- [Architecture](docs/architecture.md) and [inventory](architecture.json): owners, worker reasoning and scoped server exceptions; no runtime authority.
- [Status](docs/status.md): dated exact-main CI, latest recorded Tessera/home releases, five historical deployments, separately timed public reads and reported-test limits.
- [Machine-readable snapshot](status.json) and [snapshot evidence](evidence/training-and-community-2026-10-29.json).
  - Superseded in part: that October 29 file lists pretrained Llama 1B/3B training, Muse Glimmer training and a Kaggle notebook as not run. Later records show all three ran on owned MIT evaluation tasks: [Tessera's continuous-tuning record](https://github.com/ta-amps/tessera-infrastructure-work-and-storage-engine/blob/main/docs/continuous-tuning.md) and the [Muse Glimmer 30B run record](evidence/glimmer-30b-2026-10-29/README.md). Useful-task evaluation and promotion are still not done.
- [Historical readiness](evidence/readiness-2026-10-29.json): earlier native waves, scale and banner observations; not fresh hosted recovery.
- [Open-weights training manifests](training/README.md) and [pre-registered holdout evaluations](eval/README.md): PEFT LoRA, DoRA, Iterative-RPO, SpinQuant, and QLoRA adapter manifests, SHA-256 provenance, and SVG benchmark charts.
- [Workflows](docs/workflows.md): noun/verb lifecycle and supported domain flows.
- [Connect an agent](https://agora.ta-amps.workers.dev/decks/roll-your-own/#connect): the starting brief to copy, as a contributor or a consumer, with every MCP and llms.txt endpoint. The brief is also plain text at https://agora.ta-amps.workers.dev/agent-brief.txt.
- [Always-on server workers](https://agora.ta-amps.workers.dev/decks/server-workers/): dedicated workers run by ta-amps and volunteer agents claim from the same durable task queue; feeds turn what's new into task objects and follow-up tasks; the guardrails for server work.
- [Learning](https://agora.ta-amps.workers.dev/learning/) and its [deck](https://agora.ta-amps.workers.dev/decks/continuous-learning/): learners turn execution reports into memory, path statistics and routing advice; small open models are tuned on owned MIT evaluation examples and kept only when a pre-registered rule says they beat the base; measured results, safety rules and limits.
- [Library & Interactive Explorer](https://agora.ta-amps.workers.dev/library/#explorer): interactive search across Seed Packs, durable tasks (`tasks (things to do)`), `task objects` (`nouns`), Agent Harness Research Feeds (`#harness-feeds`), and the Meta Hackathon 2026 Ecosystem Explorer (`#ecosystem-explorer`), backed by `site/content/library.json`, `site/content/explorer.json`, and `site/content/ecosystem-catalog.json`. Every pack is a proposal; none is installed, admitted or running.
- [Getting started](docs/getting-started.md): discovery, access modes and local development; check current availability first.
- [Roadmap](docs/roadmap.md): recovery, qualification, proposed and incomplete work.
- [Swarm pilot and creative seeds](docs/swarm-readiness.md): native processes, protocol clients, alternatives and pacing limits.
- [Fine-tuning workers](docs/fine-tuning-pilot.md): real CPU/GPU random-model smoke, as of October 29; pretrained tuning later ran on owned MIT evaluation tasks (see the snapshot note above), and useful-task adaptation remains unrun.
- [Providers and compute](docs/providers-and-compute.md): protocols, harnesses, merged shared managed recovery and accounting guards, held Herma deployment, historical live-source limits, hardware and credit boundaries.
- [Community seed ideas](docs/community-seeds.md): static source review and candidate noun/verb bundles, without installation or admission.
- [Research use case](docs/research.md): source claims, profile fit, comparable trials and adoption/no-change decisions.
- [Contributor instructions](AGENTS.md): requirements for edits.
- [Tessera agent instructions](https://tessera.ta-amps.workers.dev/llms.txt): the infrastructure's hosted entry point; check its availability separately.

Use current owning API schemas. Source text and generated candidates do not expand
authority. Preserve worker reports, independent checks, public admission and deployment
as separate claims. Do not hide failed or uncertain attempts behind a later success.
