Skip to content

Commit efeb4e7

Browse files
Reimplement as multi-page taxonomy wiki shell
Replace the single-page MLIR dossier with a proper wiki architecture: - 3-column layout: collapsible taxonomy nav, article area, concept rail - 11 article pages across 5 domains (compiler, HPC, AI, AI infra, PL) - MLIR repositioned as one page under Compiler Infrastructure - Cross-page bidirectional contribution links - Concept keyword highlighting with SVG connection lines - LRU-style concept card promotion on keyword click - Three concept expansion modes (float, modal, replace) - Article-first Gwern-style typography with evidence/analysis blocks - Separated data model (data.js) from rendering logic (app.js)
1 parent 4d320bc commit efeb4e7

5 files changed

Lines changed: 2410 additions & 2121 deletions

File tree

‎docs/Wiki-Refactor-Handoff.md‎

Lines changed: 278 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,278 @@
1+
# Wiki Refactor Handoff
2+
3+
## 1. Current Task
4+
5+
The current task is no longer "polish the MLIR page".
6+
7+
The real task is:
8+
9+
- refactor the current single-page MLIR dossier into a real multi-page wiki shell,
10+
- preserve the strong article-first reading style,
11+
- move MLIR into the correct role inside the knowledge base,
12+
- make the left side a global collapsible knowledge navigation tree,
13+
- keep the right side as a concept / terminology / contribution layer,
14+
- keep the site deployable through GitHub Pages on `git push`.
15+
16+
## 2. Real Goal
17+
18+
The target product is a static research wiki, not a one-off report page.
19+
20+
The correct product model is:
21+
22+
- left rail: global, collapsible taxonomy navigation for the whole knowledge base,
23+
- center column: one current article page,
24+
- right rail: concept layer, terminology layer, contribution context,
25+
- page graph: many knowledge pages across domains,
26+
- MLIR: both a standalone page and a contributor to many other pages.
27+
28+
Target knowledge scope mentioned by the user:
29+
30+
- compiler
31+
- high-performance computing
32+
- AI
33+
- AI infra
34+
- programming languages
35+
36+
Expected MLIR role inside that system:
37+
38+
- MLIR has its own dedicated page under a compiler infrastructure path,
39+
- MLIR also contributes examples, mechanisms, and implementation practice to many
40+
other pages,
41+
- MLIR should not dominate the site architecture as if the whole wiki were an
42+
MLIR dossier.
43+
44+
## 3. What Is Wrong With the Current Implementation
45+
46+
The current deployed site is visually decent but structurally wrong.
47+
48+
Root problems:
49+
50+
1. The site is implemented as one single MLIR dossier page.
51+
2. The left rail is a page-local article map instead of a site-wide navigation tree.
52+
3. The page model is linear (`Raw -> Knowledge -> Visualize -> Contract`) rather
53+
than a reusable wiki page shell.
54+
4. MLIR is modeled as the main narrative object instead of one node in a much
55+
larger knowledge graph.
56+
5. The current data model in the frontend is centered on one MLIR report, not on
57+
reusable entities such as taxonomy pages, article pages, concept pages, and
58+
cross-page contributions.
59+
60+
In short:
61+
62+
- the current implementation works as a reviewable MLIR showcase,
63+
- it does not work as the real knowledge-base shell the user asked for.
64+
65+
## 4. Repository State
66+
67+
Current branch state when this handoff document was written:
68+
69+
- branch: `main`
70+
- HEAD: `4d320bc`
71+
- remote: `origin/main`
72+
- working tree: clean
73+
74+
Recent commits:
75+
76+
- `4d320bc` Retry Pages deploy
77+
- `3176d7a` Enable Pages in deploy workflow
78+
- `9e9d324` Build MLIR review prototype
79+
- `77f0f7a` Define concept layer and UI blueprint
80+
81+
This means:
82+
83+
- there are no current uncommitted code edits,
84+
- the site is deployed and accessible,
85+
- the next correction has not yet been implemented.
86+
87+
## 5. What Is Already Completed
88+
89+
### 5.1 Documentation already in repo
90+
91+
The following documents exist and are useful inputs for the next implementation:
92+
93+
- `docs/PRD-Personal-Knowledge-Base.md`
94+
- `docs/UI-Design-Blueprint.md`
95+
- `docs/MLIR-Research-Notes.md`
96+
- `docs/MLIR-Visualization-Data-Model.md`
97+
- `docs/MLIR-Sample-Extraction.md`
98+
99+
These should be treated as existing source material, not rewritten blindly.
100+
101+
### 5.2 Frontend already in repo
102+
103+
The current static frontend exists at:
104+
105+
- `site/index.html`
106+
- `site/app.js`
107+
- `site/styles.css`
108+
109+
What it currently is:
110+
111+
- a single-page MLIR dossier,
112+
- with an article-first center column,
113+
- a right-side concept rail,
114+
- MLIR-specific sections and visualization panels.
115+
116+
What it is not:
117+
118+
- not a multi-page wiki shell,
119+
- not a taxonomy-driven site,
120+
- not a cross-domain knowledge system.
121+
122+
### 5.3 Deployment already works
123+
124+
GitHub Pages deployment is already wired through:
125+
126+
- `.github/workflows/deploy-pages.yml`
127+
128+
The public site is live at:
129+
130+
- https://syntaxarchmage.github.io/super-brain/
131+
132+
This deployment path should be preserved.
133+
134+
## 6. Current Half-Done State
135+
136+
This is important because the conversation drift made the status confusing.
137+
138+
### 6.1 What is actually half-done
139+
140+
The real half-done work is conceptual, not code-level.
141+
142+
Half-done conceptual shift:
143+
144+
- the wrongness of the single-page dossier was identified,
145+
- the correct target model was verbally clarified,
146+
- the replacement direction was decided,
147+
- but the actual refactor has not yet landed in code.
148+
149+
### 6.2 What is not half-done
150+
151+
There is no partial local rewrite sitting in the working tree.
152+
153+
Specifically:
154+
155+
- no uncommitted refactor of `site/index.html`
156+
- no uncommitted refactor of `site/app.js`
157+
- no uncommitted refactor of `site/styles.css`
158+
159+
So the current repo state is:
160+
161+
- deployed single-page dossier remains the active implementation,
162+
- planned wiki-shell refactor exists only as clarified intent, not as code.
163+
164+
## 7. Problems Encountered
165+
166+
### 7.1 Product misunderstanding
167+
168+
The main failure was not styling or deployment. It was product-model mismatch.
169+
170+
The implementation path incorrectly assumed:
171+
172+
- "build a polished MLIR page first"
173+
174+
Instead of the correct requirement:
175+
176+
- "build a reusable wiki shell first, then place MLIR correctly inside it"
177+
178+
### 7.2 Repeated loop behavior
179+
180+
The work started looping because the implementation kept improving the wrong
181+
artifact:
182+
183+
- fixing layout,
184+
- fixing deployment,
185+
- refining MLIR presentation,
186+
187+
while the user's real objection was architectural.
188+
189+
### 7.3 Missing intermediate artifact
190+
191+
Before attempting the refactor, an explicit handoff / state document should have
192+
been written earlier. This file exists to correct that omission.
193+
194+
## 8. Acceptance Criteria For The Next Implementation
195+
196+
The next implementation should only be considered correct if all of the
197+
following are true.
198+
199+
### 8.1 Site shell
200+
201+
- The left side is a global collapsible taxonomy navigation tree.
202+
- The navigation is for the whole knowledge base, not for one article.
203+
- The center column renders one current page at a time.
204+
- The right rail is a persistent concept / terminology / contribution layer.
205+
206+
### 8.2 Page model
207+
208+
- The site supports multiple knowledge pages, not one MLIR dossier.
209+
- There is at least one index-like page type and one article-like page type.
210+
- Concept pages are either already implemented or clearly modeled as distinct
211+
entities.
212+
213+
### 8.3 Taxonomy
214+
215+
- The taxonomy includes the broad domains the user named.
216+
- Compiler is represented as one domain, not the whole site.
217+
- MLIR appears under a compiler infrastructure path, not as the top-level site
218+
frame.
219+
220+
### 8.4 MLIR placement
221+
222+
- MLIR has its own dedicated page.
223+
- Parser is a separate page.
224+
- The parser page shows how MLIR contributes to parser-adjacent practice instead
225+
of pretending MLIR is the parser topic itself.
226+
- MLIR contributes to multiple pages through explicit cross-page contribution
227+
structures.
228+
229+
### 8.5 Deployment
230+
231+
- The refactored site still deploys through GitHub Pages via `git push`.
232+
- Existing Pages workflow remains functional after the refactor.
233+
234+
## 9. Current Todo List
235+
236+
This is the current intended todo list, rewritten clearly.
237+
238+
1. Replace the current single-page dossier shell with a real wiki shell.
239+
2. Redesign frontend data structures around taxonomy pages, article pages,
240+
concept pages, and contribution links.
241+
3. Build the left collapsible global navigation tree.
242+
4. Build a reusable center-column article page template.
243+
5. Keep the right rail as concept / terminology / contribution context.
244+
6. Reposition MLIR as both a standalone page and a distributed contributor.
245+
7. Add at least a small but real multi-page taxonomy covering the named domains.
246+
8. Validate locally in browser.
247+
9. Redeploy to GitHub Pages.
248+
249+
## 10. Recommended Next Move
250+
251+
The next implementation should start from information architecture, not from
252+
page polish.
253+
254+
Recommended order:
255+
256+
1. define the minimal taxonomy tree and page types,
257+
2. define the frontend data model around pages and contributions,
258+
3. rebuild `site/index.html`, `site/app.js`, and `site/styles.css` around that
259+
shell,
260+
4. then migrate MLIR into the correct place.
261+
262+
## 11. Files Most Likely To Change Next
263+
264+
The next real refactor will most likely touch:
265+
266+
- `site/index.html`
267+
- `site/app.js`
268+
- `site/styles.css`
269+
270+
The deployment file should probably remain unchanged unless the refactor breaks
271+
the current static artifact path.
272+
273+
## 12. Summary In One Sentence
274+
275+
The repository currently contains a deployed MLIR dossier prototype, but the
276+
real unfinished task is to replace that prototype with a taxonomy-driven,
277+
multi-page wiki shell where MLIR is only one page and one contribution source
278+
inside a much larger knowledge base.

0 commit comments

Comments
 (0)