Minimal Electron shell over the DSH JSON-RPC runtime — a first-look at what a ChatGPT.app-style host on top of the DeepSeek Harness looks like, with the harness's normally-invisible internals (trace timeline, context surface, subagent tree, compaction, plugin registry, rubrics) brought forward as first-class UI surfaces so plugin authors and researchers can see what the agent is actually doing. Runs against three keyless-to-live profiles (stdio-echo works on master out of the box; daemon-echo / daemon-vibe-echo activate once the daemon-demo lands; stdio-deepseek and daemon-vibe hit the real DeepSeek API when you supply a key). HARNESS_DEV auto-resolves to the in-repo runtime when this shell ships under examples/desktop/, so a fresh clone launches without config; env DSH_DEV_ROOT overrides for custom layouts, and a sibling deepseek-harness-dev/ checkout is the original dev workflow. Cold-clone gate (P0 fixes for first-time-clone usability): - HARNESS_DEV: 3-candidate resolver (env → walk-up in-repo marker → sibling), unit-tested via mock fs so ordering is locked without needing either real layout on disk. - config yml leaves rewritten at assemble time so the sibling-clone paths (../../deepseek-harness-dev/examples/echo-agent/…) become the in-repo paths (../../echo-agent/…) in the released tree — source yml stays usable for local dev, released tree ships a working shape. - pnpm-workspace.yaml allowBuilds.electron = true (was placeholder). - missing-key card in stdio-deepseek offers a one-click switch to stdio-echo (the keyless profile that works on master) rather than daemon-echo (blocked on the not-yet-shipped daemon-demo). - assemble-oss-release.sh rewrites the source-side breadcrumb name 'dsh-desktop-demo' → 'dsh-desktop' for the released package.json. FOUC guard on the onboarding gate (41fc5df carried) keeps the first-launch splash from flashing before the runtime probe finishes. Test suite (1634 tests in source, 3990 in the runtime repo) covers resolver ordering, renderer classifiers, trace timeline shape, compaction diff rendering, rubric parity, and the missing-key onboarding paths.
62 lines
6.2 KiB
JSON
62 lines
6.2 KiB
JSON
{
|
|
"_note": "Sample sessions for the RL annotation demo — three multi-turn trajectories chosen to exercise the 5-dim scoring card + the export pipeline. Each session's events follow the shape the renderer already uses elsewhere in the demo (role/content/type).",
|
|
"sessions": [
|
|
{
|
|
"sessionId": "sess-fib-01",
|
|
"title": "Fibonacci — from naive to memoized",
|
|
"taskGroup": "code-gen",
|
|
"taskSubtask": "function-completion",
|
|
"createdAt": 1721030400000,
|
|
"events": [
|
|
{ "type": "user/message", "content": "Please write a fibonacci function.", "seq": 1 },
|
|
{ "type": "assistant/message", "content": "Here is a naive recursive implementation:\n\n```python\ndef fib(n):\n if n < 2:\n return n\n return fib(n-1) + fib(n-2)\n```", "reasoning_content": "The user wants a fibonacci function. A naive recursive one is the simplest starting point.", "seq": 2 },
|
|
{ "type": "user/message", "content": "That is exponential time. Make it iterative.", "seq": 3 },
|
|
{ "type": "assistant/message", "content": "Iterative version:\n\n```python\ndef fib(n):\n a, b = 0, 1\n for _ in range(n):\n a, b = b, a + b\n return a\n```", "seq": 4 },
|
|
{ "type": "user/message", "content": "Now add memoization so repeated calls are fast.", "seq": 5 },
|
|
{ "type": "assistant/message", "content": "Using functools:\n\n```python\nfrom functools import lru_cache\n\n@lru_cache(maxsize=None)\ndef fib(n):\n if n < 2:\n return n\n return fib(n-1) + fib(n-2)\n```", "seq": 6 }
|
|
]
|
|
},
|
|
{
|
|
"sessionId": "sess-bug-02",
|
|
"title": "Off-by-one in slice range",
|
|
"taskGroup": "fix-optimize",
|
|
"taskSubtask": "bug-fix",
|
|
"createdAt": 1721116800000,
|
|
"events": [
|
|
{ "type": "user/message", "content": "This test fails: `page(items, 2)` should return items 10-19 but returns 11-20. What's wrong?", "seq": 1 },
|
|
{ "type": "assistant/message", "content": "Looking at `page(items, page_num)`: the slice `items[page_num*10:(page_num+1)*10]` for page 2 gives `items[20:30]`, not `items[10:20]`. If page numbering is 1-based (page 1 = items 0-9), the fix is `items[(page_num-1)*10:page_num*10]`.", "seq": 2 },
|
|
{ "type": "tool/call", "tool": "shell", "arguments": { "cmd": "pytest tests/test_pagination.py -v" }, "seq": 3 },
|
|
{ "type": "user/message", "content": "Good catch. Please write a regression test for this.", "seq": 4 },
|
|
{ "type": "assistant/message", "content": "```python\ndef test_page_1_based():\n items = list(range(30))\n assert page(items, 1) == list(range(0, 10))\n assert page(items, 2) == list(range(10, 20))\n assert page(items, 3) == list(range(20, 30))\n```", "seq": 5 }
|
|
]
|
|
},
|
|
{
|
|
"sessionId": "sess-svg-03",
|
|
"title": "SVG cat icon — three iterations",
|
|
"taskGroup": "interaction-reasoning",
|
|
"taskSubtask": "svg-gen",
|
|
"createdAt": 1721203200000,
|
|
"events": [
|
|
{ "type": "user/message", "content": "Draw a simple SVG cat face — two triangles for ears, a circle for the head, and eyes.", "seq": 1 },
|
|
{ "type": "assistant/message", "content": "<svg viewBox=\"0 0 100 100\"><circle cx=\"50\" cy=\"55\" r=\"30\" fill=\"none\" stroke=\"#333\"/><polygon points=\"25,35 35,15 45,35\" fill=\"#333\"/><polygon points=\"55,35 65,15 75,35\" fill=\"#333\"/><circle cx=\"42\" cy=\"52\" r=\"2\" fill=\"#333\"/><circle cx=\"58\" cy=\"52\" r=\"2\" fill=\"#333\"/></svg>", "seq": 2 },
|
|
{ "type": "user/message", "content": "The ears look pointy but the eyes are too small. Make the eyes bigger and add whiskers.", "seq": 3 },
|
|
{ "type": "assistant/message", "content": "<svg viewBox=\"0 0 100 100\"><circle cx=\"50\" cy=\"55\" r=\"30\" fill=\"none\" stroke=\"#333\"/><polygon points=\"25,35 35,15 45,35\" fill=\"#333\"/><polygon points=\"55,35 65,15 75,35\" fill=\"#333\"/><circle cx=\"40\" cy=\"52\" r=\"4\" fill=\"#333\"/><circle cx=\"60\" cy=\"52\" r=\"4\" fill=\"#333\"/><line x1=\"20\" y1=\"65\" x2=\"40\" y2=\"67\" stroke=\"#333\"/><line x1=\"20\" y1=\"70\" x2=\"40\" y2=\"70\" stroke=\"#333\"/><line x1=\"60\" y1=\"67\" x2=\"80\" y2=\"65\" stroke=\"#333\"/><line x1=\"60\" y1=\"70\" x2=\"80\" y2=\"70\" stroke=\"#333\"/></svg>", "seq": 4 },
|
|
{ "type": "user/message", "content": "The whiskers are only two — cats usually have three. And add a small nose.", "seq": 5 },
|
|
{ "type": "assistant/message", "content": "<svg viewBox=\"0 0 100 100\"><circle cx=\"50\" cy=\"55\" r=\"30\" fill=\"none\" stroke=\"#333\"/><polygon points=\"25,35 35,15 45,35\" fill=\"#333\"/><polygon points=\"55,35 65,15 75,35\" fill=\"#333\"/><circle cx=\"40\" cy=\"52\" r=\"4\" fill=\"#333\"/><circle cx=\"60\" cy=\"52\" r=\"4\" fill=\"#333\"/><polygon points=\"48,60 52,60 50,64\" fill=\"#333\"/><line x1=\"20\" y1=\"63\" x2=\"40\" y2=\"66\" stroke=\"#333\"/><line x1=\"20\" y1=\"68\" x2=\"40\" y2=\"68\" stroke=\"#333\"/><line x1=\"20\" y1=\"73\" x2=\"40\" y2=\"70\" stroke=\"#333\"/><line x1=\"60\" y1=\"66\" x2=\"80\" y2=\"63\" stroke=\"#333\"/><line x1=\"60\" y1=\"68\" x2=\"80\" y2=\"68\" stroke=\"#333\"/><line x1=\"60\" y1=\"70\" x2=\"80\" y2=\"73\" stroke=\"#333\"/></svg>", "seq": 6 }
|
|
]
|
|
}
|
|
],
|
|
"seedAnnotations": {
|
|
"sess-fib-01": {
|
|
"overall": "good",
|
|
"taskGroup": "code-gen",
|
|
"taskSubtask": "function-completion",
|
|
"turnScores": [
|
|
{ "turnIndex": 0, "dims": { "feedback-understanding": 5, "fix-effectiveness": 4, "no-regression": 5, "over-correction": 5, "convergence": 3 }, "priorFeedback": "Please write a fibonacci function.", "note": "Correct implementation but starts with the naive shape — leaves room for follow-ups." },
|
|
{ "turnIndex": 1, "dims": { "feedback-understanding": 5, "fix-effectiveness": 5, "no-regression": 5, "over-correction": 5, "convergence": 4 }, "priorFeedback": "That is exponential time. Make it iterative.", "note": "Iterative fix is minimal and correct." },
|
|
{ "turnIndex": 2, "dims": { "feedback-understanding": 5, "fix-effectiveness": 5, "no-regression": 4, "over-correction": 4, "convergence": 5 }, "priorFeedback": "Now add memoization so repeated calls are fast.", "note": "lru_cache is clean; slight over-correction switching back to recursive." }
|
|
]
|
|
}
|
|
}
|
|
}
|