Ordivon Host
Replace the model session without replacing the Task.
A persistent agent control plane for goals, tasks, bounded cognition, candidate admission, harness assignment, effect proposals, verification, and outcomes. Model sessions and Runtime processes remain replaceable dependencies.
- Maturity
- Tested architectural core
- Built for
- Developers building long-running agents that must survive restart, provider replacement, and ambiguous completion.
- Latest proof
- One repository task completed in both Codex→Hermes and Hermes→Codex orders while three injected faults were rejected or recovered.
Question this project must answer
Which semantic state must survive when cognition, transport, and execution processes are replaced?
Working model
01Contextbounded input
→02Decisionadmitted candidate
→03Dispatchcorrelated effect
→04Outcomeverified projection
Where it stands now
Tested architectural core
Harness H1–H5 are closed: Host retained Task Attempt, Assignment generation, Run receipts, completion admission, and opaque Runtime correlation while rejecting a shared Provider lifecycle. The next bounded construction is Ordivon Harness v0 for bare model APIs.
Capability evidence
H1–H5Harness boundary closed
2/2replacement orders accepted
3first faults survived
What this project owns
Facts it must preserve and may commit.
- Goals and Tasks
- Host journal
- Cognition contexts
- Outcomes
What it leaves elsewhere
Responsibilities another owner or mature system should retain.
- No process ownership
- No model monopoly
- No Runtime mechanics
Questions that can still change the project
01↗02↗03↗04↗
testing
Can Host complete a general repository Goal without absorbing Runtime mechanics?
open
What operational surface is required before Host becomes installable rather than architectural proof?
open
What is the smallest Ordivon Harness that turns a bare model API into a verifiable Agent Run?
answered