Binary testing instructions, agent-seed bootstrap scenarios and binary smoke test

[Herklos]

Summary

Document how to test a CI-built PyInstaller binary, and add the tooling to do it: extra agent-seed bootstrap scenarios and a curl-only smoke test.

Solve the issue: (no issue)

Changelog

• Add packages/binary/BINARY_TESTING_INSTRUCTIONS.md: download the artifact, TENTACLES_URL_TAG=latest only for builds from bin_factory, run it, and a test checklist (HTTP surface, Node UI, seeded QA, debug API user actions, scheduler, journal, restart recovery)
• Add packages/binary/AGENTS.md pointing to it
• Add packages/binary/scripts/smoke_test.sh: boots the binary in an empty dir, checks the HTTP and setup/debug APIs, creates a throwaway wallet, stops and restarts it. Needs only bash and curl. Exit code 0 or 1
• agent-seed: bootstrap and all take repeatable --scenario (grid default, index, completed, lifecycle, all); seed-agent.sh bootstrap forwards the options
  • index: index automation on Seed kraken B
  • completed: a second index automation, created then stopped, so the UI has a completed automation
  • lifecycle: stops and restarts the grid automation and raises AutomationNameLostError if the name is lost
• New builders (index create, stop, restart), BootstrapScenario enum, typed errors, unit tests for builders, scenarios (fake debug API) and CLI options
• Docs: agent-seed README, ARCHITECTURE and skill

What's new?

Tested against the bin_factory Linux x64 binary (run 36476985898):

• Smoke test: 28 of 28 checks pass. It exits 1 and reports the reason when the node does not boot
• grid, index and completed scenarios work, re-running them is a no-op, and the UI shows Active 2, Completed 1
• lifecycle currently fails on purpose: after automation_restart the automation name is empty (the UI then shows OctoBot a00000). This is a node bug (restart_automation.py), not addressed here
• There is no seeded errored automation: a failing automation keeps retrying and stays running, it only becomes failed after recovery attempts are exhausted
• tools/tests and packages/node/tests/agent_seed: 101 passed. python -m tools.extended_linter --base origin/dev: 0 violations

🤖 Generated with Claude Code

https://claude.ai/code/session_014ymEjzdSYvMDqN5bJVwhjR

[Herklos]

CI note on the packages/node failure at dfc57c5: tests/functional_tests/test_start_check_and_stop_dca_2_evaluators_workflow.py::TestTriggerTaskDCATwoEvaluatorsDbosIntegration::test_trigger_task_dca_two_evaluators_selective_entry_fill_then_stop failed with Timed out waiting for stop completion ... within 15.0s (1 failed, 996 passed, 1 skipped).

This is not caused by this PR. It only adds/edits markdown (packages/binary/*.md, tools/agent_seed/README.md), and the failing test is a wall-clock timeout in a DBOS functional test under xdist. I am not changing the test or the node code here.

The push of c4f0c7a started a new CI run on the current head, which serves as the one re-run. If packages/node fails again on the same test, I will treat it as real and report it separately.

---

Generated by Claude Code

[Herklos]

CI note on the ubuntu-latestx64 - Python 3.13.x - octobot failure at 9e35cb6 (1 failed, 1709 passed, 38 skipped): tentacles/Services/Interfaces/web_interface/tests/distributions/test_default.py::TestDefaultDistribution::test_browse_all_pages_required_password_with_login. The classic web UI page /data_collector returned 500, and the server log shows sqlite3.OperationalError: disk I/O error behind it.

This is not caused by this PR: it does not touch web_interface, its tests, or any sqlite code (it changes tools/agent_seed, packages/binary, three new constants in octobot_node/agent_seed/constants.py, and docs). The error is an I/O failure on the runner while browsing pages under xdist.

I could not re-run the failed job: the re-run request was rejected with a 403 (Resource not accessible by integration), so someone with permission needs to re-run it. I will keep watching this PR, and if the same test fails again on a re-run I will treat it as real and dig into it separately.

(Correction: an earlier version of this comment said I had re-run the job. I had not.)

---

Generated by Claude Code

[GuillaumeDSM]

good for me when the branch is rebased and not with a dev merge commit