42 lines
3.2 KiB
Markdown
42 lines
3.2 KiB
Markdown
# Channel Wiki: #jarvis-jr-v1-storm
|
|
_Channel ID: 1548059565235503164_
|
|
_Created: 2026-09-11 20:05 UTC_
|
|
_Last sync: 2026-09-11 20:26 UTC_
|
|
|
|
## Purpose
|
|
Investigate and test emerging autonomous-agent practices, then turn credible findings into reusable workflows or experiments.
|
|
|
|
## Key Decisions
|
|
- Use `/opt/data/channels/jarvis-jr-v1-storm` as this channel's canonical local workspace.
|
|
- Use the private Gitea repository `home-v1-skills-code-agent/home-v1-channel-jarvis-jr-storm` as the channel skill source.
|
|
- Keep mutable channel state and detailed research out of `config.yaml`; configuration contains only the skill-loading directive.
|
|
- Treat social-media summaries as commentary until direct source material supports the attributed quote.
|
|
|
|
## Active Topics
|
|
- Agent orchestration using reusable skills, bounded loops, dependency graphs, and verification gates.
|
|
- Separating Boris Cherny's direct statements about loops and agent-managed agents from third-party “graph engineering” packaging.
|
|
- Evaluating Hermes as an agent operating environment against explicit workflow runtimes and visual builders.
|
|
|
|
## Key Context
|
|
- Local workspace: `/opt/data/channels/jarvis-jr-v1-storm`
|
|
- Skill repository: `https://gitea.lego-cloud.eu/home-v1-skills-code-agent/home-v1-channel-jarvis-jr-storm`
|
|
- Installed skill: `home-v1-channel-jarvis-jr-storm`
|
|
- Initial investigation: `/opt/data/channels/jarvis-jr-v1-storm/research/2026-09-11-boris-cherny-loops-graphs-claim.md`
|
|
- Framework evaluation: `/opt/data/channels/jarvis-jr-v1-storm/research/2026-09-11-hermes-vs-langgraph-langflow-agent-frameworks.md`
|
|
- Initial finding: Boris directly supports persistent loops/routines, large parallel agent fleets, and agents coordinating agents. The inspected transcripts did not support the screenshot's purported exact quote, its “graph engineering” attribution, or its three-to-six-month prediction.
|
|
- Framework finding: preserve Hermes as the operator-facing environment and add native durable workflow-graph semantics inspired by LangGraph, Microsoft Agent Framework, and CrewAI. CrewAI is the closest current framework-level peer because it combines event-driven flows, checkpoint/fork support, HITL, tracing, and progressive `SKILL.md` skills. Treat Langflow as a serializable visual-DAG and DevOps authoring influence, keep Agent Skills portable, and do not introduce LangGraph or Langflow as a mandatory runtime.
|
|
|
|
## People & Roles
|
|
- **RootAtSkic:** channel owner and experiment sponsor.
|
|
- **Hermes/Jarvis Jr:** investigator, implementer, and verifier within current authority.
|
|
|
|
## Action Items
|
|
- [ ] Select one existing operational workflow for a bounded loop-and-graph experiment.
|
|
- [ ] Define node contracts, retry budget, stop conditions, evidence, and human-owned gates before automating it.
|
|
- [ ] If approved, scope a native Hermes workflow-graph specification before implementation: typed state, durable node checkpoints, interrupts, idempotency, joins, bounded cycles, traceability, and exact skill provenance.
|
|
|
|
## Source Anchors
|
|
- Triggering setup and investigation request: Discord message `1548060067905933362`.
|
|
- Initial research brief: `research/2026-09-11-boris-cherny-loops-graphs-claim.md` in the local channel workspace.
|
|
- Framework comparison request: Discord message `1548063523270168617`.
|