xonecas.io
64875
Mc

mini-coder / commit 66b1e20

chore: consolidate UI testing findings into TODO.md

- Enrich existing TODO items with specifics from multi-provider UI testing
- Add status bar cwd truncation issue (new finding)
- Add Polish section for truncation, sed classification, GPT reasoning verbosity
- Update agent commands
- Remove research.md

xonecas2026-03-19 66b1e20f18ddb88a67c1940face72eda82bdc442 parent b00091f

4 files changed

.agents/commands/idea-check.md+1−1
@@ -9,7 +9,7 @@ Use the shell tool and `bun run dev` to do manual tests. One shot and interactiv
99 Employ YAGNI in your review.
1010 Check TODO.md for know issues already tracked.
1111
12-Question design choices and rethink as needed while staying compliant with the original idea. This includes changing UI and console output as needed.
12+Question design choices and rethink as needed while staying compliant with the original idea. This includes changing UI and console output as needed. Make sure all documentation matches the idea file.
1313
1414 If you think the codebase is now at a good place, check the recent commit for all changes and report the final status to the user.
1515
.agents/commands/ui-probe.md+10−31
@@ -2,15 +2,15 @@
22 description: Run one-shot mc sessions across all SDK paths, capture output, and write a UI/UX analysis report to UI-REPORT.md.
33 ---
44
5-The goal is to capture a realistic perspective of what the users see when using mini-coder.
5+The goal is to capture a realistic perspective of what the users see when using mini-coder. And ensure the core idea (mini-coder-idea.md) is well executed in the output.
66
7-Run `mc` one-shots using one cheap model per Zen SDK path, capture their raw terminal output, then analyse it yourself and write `UI-REPORT.md` in the repo root.
7+Run `mc` using one cheap modern model per provider path, capture their raw terminal output, then analyse it yourself and write `UI-REPORT.md` in the repo root.
88 Do NOT delegate the report to the models under test.
9-Use `tmux` so you can test interactive sessions properly.
9+Use `tmux` so you can test interactive sessions properly. Remember to send Enter to submit, and careful with the paste handling.
1010
1111 ## Models to use (one per SDK path)
1212
13-Pick cheap options. Verify available models at https://opencode.ai/docs/zen/ if needed, then choose:
13+Pick cheap options. Verify available models at https://opencode.ai/docs/zen/, then choose:
1414
1515 - `@ai-sdk/anthropic` path → a `zen/claude-*` haiku-class model
1616 - `@ai-sdk/google` path → a `zen/gemini-*` flash-class model
@@ -22,38 +22,17 @@ Also use a Anthropic Oauth model for completness
2222
2323 ## Prompt to give each model
2424
25-```
26-Read mini-coder-idea.md and TODO.md, then create mc-summary.md in the repo root with a short summary of what mini-coder is and a list of the key TODO items.
27-```
28-
29-This prompt is intentionally simple: it covers file reads, a write, and exposes path resolution, spinner, error handling, and output formatting behaviour.
30-
31-## Steps
32-
33-1. Create `logs/` in the repo root if it doesn't exist.
34-2. For each model, run:
35- ```
36- mc -m <model> "<prompt>" > logs/<model>.log 2>&1
37- ```
38- Run them sequentially (not in parallel) to avoid SQLite lock issues.
39-3. Strip ANSI escape codes from each log for clean reading:
40- ```
41- sed 's/\x1b\[[0-9;?]*[mKlh]//g; s/\r//g' logs/<model>.log > logs/<model>-clean.log
42- ```
43-4. Read all clean logs yourself and analyse what a user would actually see.
44-5. Clean up any `mc-summary.md` files left behind by the test models. And the log files
25+Use prompts that will ensure a realistic pattern, using multi steps and turns. Check the db for examples if needed.
4526
4627 ## Report structure for UI-REPORT.md
4728
4829 Write `UI-REPORT.md` in the repo root with these sections:
4930
50-- **Methodology** — models tested, prompt used, how output was captured
51-- **Per-model results** — for each model: did it complete the task? what did the user see step by step? any errors or recovery detours?
52-- **Shared UI observations** — patterns across all models: spinner behaviour, tool call chrome (symbols, indentation), reasoning block display, diff output, final response format, banner/status bar presence
53-- **UI/UX** how can it be improved, is the output elegant, clear and hierchial?
54-- **Bugs found** — table: severity, description, likely source file
55-- **Alignment with mini-coder-idea.md** — table comparing each idea goal against what was observed (✅ / ⚠️ / ❌)
56-- **Recommendations** — split into immediate bugs and polish items
31+- **Summary** -- a short summary of how you tested and what your found.
32+- **Alignment with mini-coder-idea.md** -- Comparison summary, avoid using tables.
33+- **Recommendations** -- split into immediate bugs and polish items
34+
35+Don't be verbose, be concise in your report.
5736
5837 Use `mini-coder-idea.md` and `TODO.md` as the source of truth for what the UI _should_ look like.
5938 The report is analysis only — no code changes.
TODO.md+11−1
@@ -1,8 +1,18 @@
11 # TODO
22
3-- CTRL+d hangs sometimes and doesn't exit
3+- Bug: Users cannot resume sessions, `mc -r <sessiond id>` starts a new session instead of resuming correctly.
4+- Prompt UI is missing provider information in some cases. Zen models omit the `zen/` prefix while OAuth models correctly show `anthropic`. Refactor so current provider is always clear.
5+- Banner doesn't show reasoning or verbose status. Users can't tell if these are active without toggling.
6+- Status bar drops the cwd path when the line gets long (observed with long OAuth model names on second turn). May be a terminal width truncation issue.
7+- Model DB sync doesn't include newer Zen models. `gpt-5.4-nano`, `gpt-5.4-mini`, `gpt-5.4` are available on Zen but absent from `provider_models`. The sync may be running against a stale endpoint or filtering too aggressively.
8+- CTRL+d hangs sometimes and doesn't exit.
49 - Codebase is hurting from fast iteration and adding/removing features as well as core fundamental changes.
510
11+## Polish
12+
13+- **Truncated tool commands** — Long `mc-edit` invocations truncate with `…` mid-argument (e.g. `--new '{…`). Could truncate at argument boundaries for clarity.
14+- **`sed` classification in shell icons** — `sed -n '1,200p'` (read-only) shows the `✎` write icon. Complex pipes like `grep … | sed …` show `?`. Could check for `-n`/`-i` flags, or classify `sed` as `$` (generic run).
15+
616 ## Deferred fixes
717
818 - `/_debug` hidden command that snapshots recent logs/db and creates a report in the cwd. For dev mostly but available to all. Do not list it anywhere, only documented here and in the code itself.
research.md+0−38
@@ -1,38 +0,0 @@
1-I want your help exploring how could we use yoctocolor to add color ouput to commonly used shell commands. Let's first check the db, check the recent past sessions for a sample. Then check for the shell usage and assert the most common commands
2-
3-Response:
4-
5-## Shell Usage Analysis (from 101 sessions)
6-
7-Most common commands:
8-| Command | Count | Opportunity |
9-|---------|-------|-------------|
10-| `cat` | 72 | File content preview |
11-| `git` | 54 | Status/diff output |
12-| `bun` | 32 | Test/build output |
13-| `sed` | 29 | Transform preview |
14-| `mc-edit` | 26 | File edits |
15-| `nl` | 12 | Line numbers |
16-| `rg` | 11 | Search matches |
17-| `npm` | 9 | Package output |
18-| `grep` | 7 | Match highlighting |
19-| `ls` | 5 | File listings |
20-
21-## Current Shell Output Coloring (in `tool-result-renderers.ts`)
22-
23-The shell result renderer already uses yoctocolors:
24-
25-- **Status badges**: `c.green("done")` / `c.red("error")` / `c.yellow("timeout")`
26-- **Stderr lines**: `c.red()` for error emphasis
27-- **Stdout/stderr line separators**: `c.dim()` or colored pipe character
28-
29-## Enhancement Opportunities
30-
31-1. **Command syntax highlighting** - Color `git`, `bun`, `npm` commands differently in the tool call line
32-2. **File path highlighting** - Use `c.cyan()` for paths in output
33-3. **Git diffs from mc-edit** - Render a colored diff
34-4. **Line numbers** - `nl` output could use `c.yellow()` for numbers
35-5. **Git output** - Branch names in `c.cyan()`, commit hashes in `c.yellow()`, file statuses (M/A/D) in color
36-6. **rg/grep matches** - Highlight matching text in `c.greenBright()` or `c.yellow()`
37-7. **Exit code** - Non-zero codes in `c.red()` or `c.yellow()`
38-8. **Test output** - `bun test` results: passes in `c.green()`, fails in `c.red()`, skips in `c.yellow()`