Skip to content

ReadyAgents Core 0.2.0

Choose a tag to compare

View all tags
@readyagents readyagents released this 19 Aug 12:15
· 172 commits to main since this release
v0.2.0
21e8ab3

ReadyAgents Core 0.2.0

The first release that feels like a toolkit you can trust for one-shot work.

ReadyAgents Core is still the free, open-source engine: YAML workflows, BYOK, builtin tools, MCP optional, packs later. 0.1.0 proved clone-and-run. 0.2.0 is the reliability and operator-experience layer.

Apache-2.0. No vendor keys. No always-on runtime in core.

Why this release matters

A careful engineer running agent graphs needs four things a playground does not:

  1. A human can stop a run. Approval nodes pause (they do not hang a TTY).
  2. A failed step is not a lost run. State is written after every node; resume continues from the last success.
  3. You can see what happened. runs list / show / inspect / replay.
  4. Starting a project is boring in the good way. readyagents new writes a workflow, README, and .env.example.

That is the difference between a demo engine and something you would actually use.

What is new

Human-in-the-loop

- id: gate
  type: approval
  prompt: "Release payment of {{total}}?"
  then: receipt
  else: denied

Without a decision the CLI exits 2, persists paused, and tells you how to continue:

readyagents run examples/approval_gate.yaml
readyagents resume <run_id> --approve gate
# or
readyagents run examples/approval_gate.yaml --approve gate

Durable local runs

JSON under .readyagents/runs/<id>.json, written atomically after each node.

readyagents runs list
readyagents runs show <run_id>
readyagents runs show <run_id> --json
readyagents runs replay <run_id>
readyagents resume <run_id> --approve gate

Filters: --status, --workflow, --limit. Logs include run= and node=.

Composition

  • type: parallel — independent branches concurrently (examples/fanout_gate.yaml)
  • type: include — run another workflow file (examples/include_demo.yaml)

Scaffolding

readyagents new my-flow                     # approval template (default)
readyagents new my-flow --template basic
readyagents new my-flow --template research # parallel + approval

Each writes workflow.yaml, README.md, and .env.example.

Dry-run

Agent nodes report usage: estimated_tokens=…. LLM examples that parse_json after an agent no longer crash on the dry-run stub.

Try it (no API keys)

pip install -e .
readyagents run examples/calc_pipeline.yaml
readyagents runs list

readyagents run examples/approval_gate.yaml --approve gate
readyagents run examples/fanout_gate.yaml --approve gate
readyagents run examples/include_demo.yaml

readyagents new demo --template research
readyagents run demo/workflow.yaml --approve publish

With keys, the existing research / triage / review examples still work. --dry-run still walks them without calling a vendor.

What we deliberately left out of core

These belong in waitlisted packs, not for sale:

  • Always-on / continuous workers and schedulers
  • Hosted control plane, teams, SSO
  • Distributed recovery and remote run stores
  • Alerting, paging, billing

Core stays small so those can plug in via readyagents.packs.

Compatibility

  • Python 3.11+
  • Existing 0.1.0 workflow YAML still runs. New node types (approval, parallel, include) are additive.
  • CLI exit code 2 now means “paused for approval”, not a generic failure.

Docs

Assets 2
Loading