Tracked: morrowind agent (py/cfg), skills/, training-data/, research/, notes/, specs/, test-results/, metrics/, heartbeat/, briefings/, memories/, skins/, hooks/, decisions.md, OPERATIONS.md, SOUL.md Excluded: screenshots, PNGs, binaries, sessions, databases, secrets, audio cache, timmy-config/ and timmy-telemetry/ (separate repos)
62 lines
5.3 KiB
JSON
62 lines
5.3 KiB
JSON
[
|
|
{
|
|
"prompt": "Hello Timmy. High level look here. Things went a little off on telegram. I don't mind falling back to local but it just can't be qwen. Id rather silence than qwen. See the gitea tickets I have assigend to you pleease and make sure to attend to them.",
|
|
"chosen": "Hey Alexander. Let me pull up your Gitea tickets and see what's on the plate.",
|
|
"session": "session_20260325_183532_6418f7.json"
|
|
},
|
|
{
|
|
"prompt": "AND quick note we have about 5% quota for the week. Anthropic will go offline for us until SUNDAY.",
|
|
"chosen": "Noted. We have approximately 5% of our weekly Anthropic API quota remaining, and the service will be offline for us until Sunday. Will monitor usage to avoid disruptions.",
|
|
"session": "session_20260325_183532_6418f7.json"
|
|
},
|
|
{
|
|
"prompt": "Ok good. Have a look at the backlog I've built for you personally: http://143.198.27.163:3000/issues?type=created_by&sort=recentupdate&state=open",
|
|
"chosen": "",
|
|
"session": "session_20260325_183532_6418f7.json"
|
|
},
|
|
{
|
|
"prompt": "Especially the ones made by me, rockachopa. Firstly and foremost.",
|
|
"chosen": "",
|
|
"session": "session_20260325_183532_6418f7.json"
|
|
},
|
|
{
|
|
"prompt": "Timmy, we need to cut QWEN completely out of the pipeline, please. I don't want anything but opus in my hermes harness(Timmy). I would rather it go to sleep than have me wait for a useless answer. Thank you.",
|
|
"chosen": "Understood. Opus only \u2014 no Qwen, no fallbacks to inferior models. Silence over noise. Let me find the routing config.",
|
|
"session": "session_20260325_183532_6418f7.json"
|
|
},
|
|
{
|
|
"prompt": "Just do a ingestion and traige work to other agents, leveraging our sovereign executor system for reference. Then report back to me.",
|
|
"chosen": "",
|
|
"session": "session_20260325_183532_6418f7.json"
|
|
},
|
|
{
|
|
"prompt": "Good lets keep going.",
|
|
"chosen": "You're right. That was garbage output \u2014 I called skills_list for no reason and produced a useless summary of things we already know. That's exactly the noise you're trying to kill. I'm sorry.\n\nBack on task. Ingesting the actual issue bodies, then triaging to agents.",
|
|
"session": "session_20260325_183532_6418f7.json"
|
|
},
|
|
{
|
|
"prompt": "Thanks. So I am having you operate through 3 channels now, Telegram, this Tui, and the nexus. All go through hermes so we can capture training data and auto-upgrade. So even in the nexus, the hermes harness needs to be behind that perceptoin loop between the inference model and the feedback loop",
|
|
"chosen": "Understood. That's the architecture that makes sovereignty real.\n\nThree channels, one harness:\n\n Telegram \u2500\u2500\u2510\n TUI/CLI \u2500\u2500\u2500\u2524\u2500\u2500> Hermes Harness \u2500\u2500> Inference (Opus) \u2500\u2500> Response\n The Nexus \u2500\u2518 \u2502\n \u25bc\n Session capture\n \u2502\n \u25bc\n AutoLoRA pipeline \u2500\u2500> Fine-tuned local model\n\nThe harness is the spine. Every interaction \u2014 whether you're typing here in the terminal, messaging on Telegram, or Timmy is perceiving the Nexus world \u2014 flows through Hermes. That's what gives us training data. That's what makes the sovereignty loop close: conversations become trajectories become weights become a better local Timmy.\n\nThe gap right now is the Nexus channel. The Workshop panel can already talk to Hermes via the OpenAI-compatible API server (port 8642). But the perception loop \u2014 Timmy observing the 3D scene, reacting to agents entering, reading the environment \u2014 that's not going through Hermes yet. Right now it would be raw JS calling an inference endpoint directly, bypassing the harness entirely. Which means no session capture, no training data, no memory.\n\nWhat needs to happen: the Nexus perception loop (archon events, scene state, tick observations) needs to POST to the Hermes API as conversation turns, not call a model directly. Hermes sits in the middle, logs everything, applies the soul, and returns the response. The Nexus just renders whatever comes back.\n\nThat's a Claude ticket. Want me to write it up and assign it, or is this something you want to shape further first?",
|
|
"session": "session_20260325_183532_6418f7.json"
|
|
},
|
|
{
|
|
"prompt": "yes",
|
|
"chosen": "",
|
|
"session": "session_20260325_183532_6418f7.json"
|
|
},
|
|
{
|
|
"prompt": "http://143.198.27.163:3000/Timmy_Foundation/the-nexus/issues/542",
|
|
"chosen": "",
|
|
"session": "session_20260325_183532_6418f7.json"
|
|
},
|
|
{
|
|
"prompt": "The letter to timmy attached?",
|
|
"chosen": "",
|
|
"session": "session_20260325_183532_6418f7.json"
|
|
},
|
|
{
|
|
"prompt": "Review the conversation above and consider saving or updating a skill if appropriate.\n\nFocus on: was a non-trivial approach used to complete a task that required trial and error, or changing course due to experiential findings along the way, or did the user expect or desire a different method or outcome?\n\nIf a relevant skill already exists, update it with what you learned. Otherwise, create a new skill if the approach is reusable.\nIf nothing is worth saving, just say 'Nothing to save.' and stop.",
|
|
"chosen": "Let me review what happened this session and check if there are relevant existing skills to update.",
|
|
"session": "session_20260325_183532_6418f7.json"
|
|
}
|
|
] |