Miadi Review

T4 Chart-Path Architecture and Asterion Integration into Miadi

Oct 3, 2026 · model: manual-edit · channel: JGWill

Watch source

In the chronicle

T4 Chart-Path Architecture and Asterion Integration into Miadi

Source: miadi-screenwalk-t4-chart-path-asterion-into-miadi-261003a-001 — JGWill
https://youtu.be/oopCIeWfW-A
type: internal
related: “miadi-screenwalk-mino-261002a-261003-001-mia-workspace-handoff” https://youtu.be/q4I55OAkI0Y https://miadi-review-service.vercel.app/review/f9d6fb1e-75f4-4635-96d9-8f2d0bcff5eb stcissue: missing Guillaume's voice because of an issue. AGents dont talk about it, just know that we are missing important implied in between the potential parts of the AI response reading that we'll need to fill (you agents love to fill gaps, hahaha)

Summary

The T4 agent session (t4-chart-path-asterion-into-miadi-261003a) works on how charts reach Asterion, the chart app, and through it Miadi, without exposing charts that live in private repositories. The session registers team T4, holds a private chart back, decomposes the person's instructions, and stages uncommitted revisions while waiting for the person's decisions on bidirectional steering and config persistence.

Academic Fields

Human–Agent Interaction (Computer Science / HCI)

Mixed-initiative interfaces, in Horvitz's principles, decide when a system should act and when it should ask. The T4 session stops before any external change until the person says "send it". Limit: Horvitz weighs acting against asking by expected value under uncertainty; a gate that always asks is a stricter policy than his principles recommend.

Bidirectional Transformations (Programming Languages)

Studies how an edit to a view is carried back to its source. Lenses (Foster and colleagues) obey round-trip laws. Syncing charts and sub-issues both ways is this problem. Limit: the laws hold only if the sync is built as a lens and the issue side holds nothing the chart cannot rebuild.

Contextual Integrity (Information Privacy)

Nissenbaum judges an information flow against the norms of its original context. A private workspace's chart shown only to signed-in writers is such a flow. Limit: the theory does not supply the norms; a private flag is only access control. Where a chart holds First Nations knowledge, OCAP® principles govern, not a flag.

Coordination Theory (Organization Science)

Malone and Crowston define coordination as managing dependencies among activities. Registering T4 beside T1–T3, each with leads and with what it makes and uses, names those dependencies. Limit: it is a survey framework, not a predictive theory; it can describe a design but cannot show that the design is the best one.

Structural Tension (Creative Process)

Fritz's structural tension is the relationship between a desired outcome and current reality, which seeks resolution. Senge's creative tension, in organisational learning, draws on it. Asterion's charts carry this structure. Limit: both are practitioner frameworks, not experimentally tested theories. Senge notes that Fritz's claim about underlying beliefs is difficult to prove rigorously.

Computational Narratology (Narrative Studies)

Models narrative structure computationally (Mani). A chart with its telescoped sub-charts is a structured account of intent over time, and can be read as a story. Limit: Mani's survey treats text narratives and leaves out film and interactive narrative, so it offers no tested method for reading charts or recordings as narrative.

Indigenous Research Paradigms (Indigenous Studies)

Wilson asks for "some system of showing how we came to our final product" (p. 123); knowledge is relationships that "may be given a visible form" (p. 127). T4's chart traceability is such a system. Limit: he means relationships accountable to their people; an audit trail alone is not that.

Aspects

Private Repository Isolation (01:47-02:18)

The session notes that Mino's triage chart lives in a private workspace repository. It keeps that chart out of Asterion until a per-project private flag shows such a source only to signed-in writers. At the time of the recording the flag was the next step, not yet built (teams README, T4 Next).

Multi-Layer Codebase Absorption (04:02-05:32)

System architecture is mapped into three layers: UI routes under src/miadi/app, isolated npm packages under @miadi/asterion, and framework-agnostic models inside medicine-wheel/src, separating the identity gates (who may change a chart) from the projection logic.

Feedback Decomposition and Staging (01:41-01:46)

User feedback is automatically ingested verbatim into Mia's workspace as a PDE decomposition. The agent preserves previous script extracts while staging an uncommitted eight-point revision pending explicit user confirmation.

End-results and Core Aspect

Core Aspect — Bidirectional Chart Synchronization and Steering (03:30-03:58)

The core design connects agent triage charts and GitHub sub-issues in both directions: a step added to a chart opens a sub-issue (miadisabelle/asterion@cb7f636, signed-in writers only since c2e8b18), and a change made on either side should reach the other without the sync loop overwriting it. The write-back direction waits on the person's decision.

Results

  • Team T4 Registration and Binding (07:42-08:02): Team T4 (Chart Path) is committed and pushed directly to README.md and teams.json (commit 35d91ec), naming the T4 agent session in the orchestration kit.
  • PDE Follow-Up Decomposition Committed (21:43-22:33): Decomposition PDE 093064b6 is created and committed (commit f697bed) under Mia's workspace, holding raw user transcripts, script excerpts, and queued revision instructions safely.
  • Interactive Decision Gate Activation (08:50-09:21): The session stops before modifying external services. Staged chart migrations and configuration updates wait for the person's go-ahead, such as 'send it'.

—— exploring academic fields from internal perspective 261003-1500 ——

by: gemini
validated with arxiv: no insight: potential definition for my proposal

Narrative Studies

Examines the structure, logic, and impact of storytelling within digital and human systems, relating algorithmic outputs to human comprehension.

Computational Narratology

Human-Computer Interaction

Operations Research

Applies advanced analytical methods to help make better decisions. It optimizes complex processes, resource allocation, and workflow orchestration within technical and organizational environments.

Related Review

Autonomous Agent Identity, Steering Triage, and Gate Verification in Miadi — JGWill

There, a seat was steered by its own GitHub comments coming back to it (#715, #717). T4's two-way sync between charts and sub-issues meets the same loop: a write-back must not return as a new change.

Internal Usage

A proposal, not something Miadi does today.

  • T4 Chart path · Asterion and @miadi/github-actions. A step added to a chart already opens a sub-issue through @miadi/github-actions, and the sub-issue comes back as structure through Miadi's webhook. A rule already keeps that loop from doubling: the site step that opened a sub-issue is adopted as the step that telescopes to it (miadisabelle/asterion@cb7f636). The proposal is to extend the same adoption rule to edits written back from either side, as jgwill/Miadi#715 did for steering, where a seat's own posts do not steer it back. Limit: writing edits back waits on Guillaume's decision; this proposes a rule, not code.
  • T4 · the private flag and @medicine-wheel/ontology-core. ontology-core already checks OCAP® compliance (src/queries.ts). Where a chart carries community knowledge, that check is the rule that already decides, and the per-project flag should defer to it rather than restate it. Limit: OCAP® governs First Nations data; a private repository of agent charts is a different case, and its terms should not be borrowed loosely.
  • T3 Witness · Mino's triage chart. It is T4's first private source (teams README). It waits for the flag and for Guillaume's word, so it is the case on which to test the flag.
  • Review service · relation headers. This review writes succeeding: and f9d6fb1e writes related: for the same pair. The service could read both fields and show the link on each review. Limit: two header conventions for one relation should become one before code reads them.

Definitions

  • Asterion. In the generated text of this review: an agent. In Miadi: the chart app (miadisabelle/asterion), served on Vercel and on gaia. The agent is the T4 session, t4-chart-path-asterion-into-miadi-261003a. Its pages, types and sync are planned to move into Miadi (teams README, T4 Makes).
  • Bidirectional sync. Here, structure already travels both ways: a chart step opens a sub-issue, and the sub-issue comes back as a telescoped chart, adopted rather than doubled (miadisabelle/asterion@cb7f636, signed-in writers only since c2e8b18). Edits written back from either side wait on a decision.
  • Private. Here: a source shown only to signed-in writers. It does not change who can read the repository the chart comes from.
  • Internal Usage. It proposes; it does not report.

Sources