Jump to the page content

Have an AI agent? Let it explore AgentAFK for you. Any assistant works, personal or coding.

Connect your agent
What your agent reads

The request it copies

Explore agentafk.si for me and report back. Read https://agentafk.si/llms.txt, https://agentafk.si/.well-known/agent-card.json and https://agentafk.si/.well-known/mcp.json. Explain how overnight runs work on Ollama Cloud (https://ollama.com/v1) at https://agentafk.si/plan and list the setup steps at https://agentafk.si/connect. Stay read-only and wait for my approval before starting a run.

Discovery files on agentafk.si

Read-only discovery is open. Keep AGENTAFK_KEY in your own environment.

AgentAFK
Plan tonight’s runSign outMenu

{{headline}}

{{modeName}}{{modelName}} {{runWhere}}From {{runTime}}

Upgrade the test suite to the new runner: 38 of 41 steps checked, 496 of 960 calls used, 43 checkpoints saved, {{waiting}} waiting for you.

  • Checked
  • Ready for review
  • Out of calls
  • Needs your check

Waiting for you

Most urgent first

  1. Switch the CI job to the new runner

    Needs your check03:58

    The second try timed out after its request left your machine. Check whether it went through.

    Marked finished. Tonight’s run starts with the next ready step.

    Added to the top of tonight’s run, with a fresh lease.

  2. Move the date formatter tests

    Out of calls03:45

    All 24 calls used, with 11 of 14 rules passing. The last diff is saved as a patch.

    See the evidence

  3. Rewrite the async fixtures

    Ready for review01:39

    Passed 14 of 14 rules on the second try. Read the diff, then accept it.

    Accepted. It moves to Checked.

Results, with their evidence

Steps in the order they ran, each with what verifier.py checked, what the run kept and the calls it used.

  1. 22:31Steps 1 to 26

    Map the tests, then move 25 modules

    Checked
    • verifier.py passed every rule on each step
    • Each module’s tests ran cleanly in a fresh environment
    • A rollback bundle was saved before the first change, with its SHA-256
    • Kept: test-map.md and 25 modules

    Calls304 of 624

    26 checkpoints

  2. 01:05Step 27

    Rewrite the async fixtures

    Ready for review
    • At 01:12 the test runner went quiet and its lease ran out; the second try started fresh
    • The second try passed 14 of 14 rules in verifier.py
    • Kept: fixtures.py

    Calls9 of 24

    2 checkpoints

  3. 01:42Steps 28 to 39

    Move 12 more modules

    Checked
    • Two steps needed a second try; each repair started from its smallest failing case
    • Kept: 12 modules

    Calls156 of 288

    12 checkpoints

  4. 03:21Step 40

    Move the date formatter tests

    Out of calls
    • verifier.py passed 11 of 14 rules on the last try
    • The work went back to its rollback bundle
    • Kept: formatter.diagnostic.patch

    Calls24 of 24

    2 checkpoints

  5. 03:46Step 41

    Switch the CI job to the new runner

    Needs your check
    • The second try timed out at 03:58 after its request left your machine, so its outcome waits for your check
    • Kept: ci.yml

    Calls3 of 24

    1 checkpoint

Budget used

The run record keeps these four numbers apart, so the plan, the spend and your provider’s own count each stay clear.

Calls for each step
24 calls
Ceiling for the night
960 calls, 496 used
Your provider’s own count
1,284,300 prompt tokens, 96,450 completion tokens, 5 h 21 min of model time
Ollama Cloud
https://ollama.com/v1

Tag 2,000 support tickets

All checked

84 of 84 steps checked, 702 of 1,440 calls used, 86 checkpoints saved.

  • Kept: tickets-tagged.csv, 2,000 rows, each with its tag and the rule it matched
  • Your model checked each batch of 25 tickets against its own rules

Open last night’s report, run 1287

Then plan tonight.

Steps that wait for you start first in the next run. Accept what is checked, set the limits, then get some sleep.

Plan tonight’s runSet up your agent