Datasets:
id stringlengths 36 36 | use_case stringclasses 2
values | instruct_type stringclasses 2
values | instruction stringclasses 2
values | input stringlengths 1.25k 2.28k | output stringlengths 67 216 | state unknown | questions unknown | answers unknown | provider stringclasses 2
values | model stringclasses 2
values | scope stringclasses 2
values | memory_id stringlengths 36 36 ⌀ | source stringclasses 1
value | created_at int64 1,790B 1,790B |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
cddc3e4f-cd2d-4f3b-a186-9fbfd5c5e1fe | analyze | choice_noul_score | Given the state (a memory to store), define and answer the TypeSafe System One questions that classify it: pick its type (choice), judge whether it is durable (noul), and score its salience (score). | {"state":{"memory":"DGUI-HyperMem syncs every JEV decision to the HuggingFace dataset ctaxnagomi/DGUI_HYPERMEM-JEV on an hourly schedule."},"questions":{"memory_type":{"type":"choice","instructions":"Which single category best describes this memory?","criteria":{"fact":"A durable statement of fact about the user, their... | {"memory_type":{"choice":"solution","probabilities":null},"durable":{"noul":0.8},"salience":{"score":2}} | {
"memory": "DGUI-HyperMem syncs every JEV decision to the HuggingFace dataset ctaxnagomi/DGUI_HYPERMEM-JEV on an hourly schedule."
} | {
"memory_type": {
"type": "choice",
"instructions": "Which single category best describes this memory?",
"criteria": {
"fact": "A durable statement of fact about the user, their environment, tools, or the world.",
"preference": "A stated like, dislike, style, or way of working.",
"decis... | {
"memory_type": {
"choice": "solution",
"probabilities": null
},
"durable": {
"noul": 0.8
},
"salience": {
"score": 2
}
} | workers-ai | @cf/meta/llama-3.1-8b-fast-v2 | dgui-hypermem | b1edc9bf-0f64-4682-816d-87962bd6674f | null | 1,789,797,335,155 |
cffacc1b-e214-4651-97a8-1cbb50f7118d | analyze | choice_noul_score | Given the state (a memory to store), define and answer the TypeSafe System One questions that classify it: pick its type (choice), judge whether it is durable (noul), and score its salience (score). | {"state":{"memory":"The JEV layer in DGUI-HyperMem records three kinds of training rows: analyze, rerank, and supersede."},"questions":{"memory_type":{"type":"choice","instructions":"Which single category best describes this memory?","criteria":{"fact":"A durable statement of fact about the user, their environment, too... | {"memory_type":{"choice":"fact","probabilities":null},"durable":{"noul":0.9},"salience":{"score":2}} | {
"memory": "The JEV layer in DGUI-HyperMem records three kinds of training rows: analyze, rerank, and supersede."
} | {
"memory_type": {
"type": "choice",
"instructions": "Which single category best describes this memory?",
"criteria": {
"fact": "A durable statement of fact about the user, their environment, tools, or the world.",
"preference": "A stated like, dislike, style, or way of working.",
"decis... | {
"memory_type": {
"choice": "fact",
"probabilities": null
},
"durable": {
"noul": 0.9
},
"salience": {
"score": 2
}
} | workers-ai | @cf/meta/llama-3.1-8b-fast-v2 | dgui-hypermem | 7e619d25-de5d-4d8c-995a-93c4795e3e43 | null | 1,789,797,337,595 |
925678d7-d729-462a-af72-a1e4ac8424cd | analyze | choice_noul_score | Given the state (a memory to store), define and answer the TypeSafe System One questions that classify it: pick its type (choice), judge whether it is durable (noul), and score its salience (score). | {"state":{"memory":"HyperMem recall is hybrid: Vectorize ANN plus D1 FTS5 keyword search, fused by reciprocal rank then re-ranked by JEV."},"questions":{"memory_type":{"type":"choice","instructions":"Which single category best describes this memory?","criteria":{"fact":"A durable statement of fact about the user, their... | {"memory_type":{"choice":"solution","probabilities":null},"durable":{"noul":0.8},"salience":{"score":3}} | {
"memory": "HyperMem recall is hybrid: Vectorize ANN plus D1 FTS5 keyword search, fused by reciprocal rank then re-ranked by JEV."
} | {
"memory_type": {
"type": "choice",
"instructions": "Which single category best describes this memory?",
"criteria": {
"fact": "A durable statement of fact about the user, their environment, tools, or the world.",
"preference": "A stated like, dislike, style, or way of working.",
"decis... | {
"memory_type": {
"choice": "solution",
"probabilities": null
},
"durable": {
"noul": 0.8
},
"salience": {
"score": 3
}
} | workers-ai | @cf/meta/llama-3.1-8b-fast-v2 | dgui-hypermem | e3151b69-0132-4155-b317-bbe7e1cadcf9 | null | 1,789,797,339,556 |
c535f66a-85b3-4942-a327-13ebe5fa402c | analyze | choice_noul_score | Given the state (a memory to store), define and answer the TypeSafe System One questions that classify it: pick its type (choice), judge whether it is durable (noul), and score its salience (score). | {"state":{"memory":"The DGUI_HYPERMEM-JEV dataset is the training brain for DGUI-HyperMem and grows with live usage."},"questions":{"memory_type":{"type":"choice","instructions":"Which single category best describes this memory?","criteria":{"fact":"A durable statement of fact about the user, their environment, tools, ... | {"memory_type":{"choice":"fact","probabilities":null},"durable":{"noul":0.8},"salience":{"score":2}} | {
"memory": "The DGUI_HYPERMEM-JEV dataset is the training brain for DGUI-HyperMem and grows with live usage."
} | {
"memory_type": {
"type": "choice",
"instructions": "Which single category best describes this memory?",
"criteria": {
"fact": "A durable statement of fact about the user, their environment, tools, or the world.",
"preference": "A stated like, dislike, style, or way of working.",
"decis... | {
"memory_type": {
"choice": "fact",
"probabilities": null
},
"durable": {
"noul": 0.8
},
"salience": {
"score": 2
}
} | workers-ai | @cf/meta/llama-3.1-8b-fast-v2 | dgui-hypermem | 4f19502a-9e37-4d24-92cc-7a3ba5cd0579 | null | 1,789,797,341,405 |
8c2c9289-8fda-490d-929c-e1b8dedb5f00 | rerank | noul | Given a query and candidate memories, evaluate for every candidate (as a noul question) whether it directly answers the query or is essential context for it. | {"state":{"query":"How does HyperMem train its JEV reasoning layer?","candidates":[{"index":0,"memory":"The DGUI_HYPERMEM-JEV dataset is the training brain for DGUI-HyperMem and grows with live usage."},{"index":1,"memory":"The JEV layer in DGUI-HyperMem records three kinds of training rows: analyze, rerank, and supers... | {"c0":{"noul":0},"c1":{"noul":0.8},"c2":{"noul":0},"c3":{"noul":0}} | {
"query": "How does HyperMem train its JEV reasoning layer?",
"candidates": [
{
"index": 0,
"memory": "The DGUI_HYPERMEM-JEV dataset is the training brain for DGUI-HyperMem and grows with live usage."
},
{
"index": 1,
"memory": "The JEV layer in DGUI-HyperMem records three kinds... | {
"c0": {
"type": "noul",
"instructions": "Memory candidate 0 is printed below under candidates[0]. Does it directly provide the information the query asks for, or essential context needed to act on it?",
"criteria": {
"true": "The candidate answers the query, or is necessary context for it, even if... | {
"c0": {
"noul": 0
},
"c1": {
"noul": 0.8
},
"c2": {
"noul": 0
},
"c3": {
"noul": 0
}
} | workers-ai | @cf/meta/llama-3.1-8b-fast-v2 | dgui-hypermem | null | null | 1,789,797,343,656 |
4a14ad29-eed1-48e1-8814-f0243360c5bd | rerank | noul | Given a query and candidate memories, evaluate for every candidate (as a noul question) whether it directly answers the query or is essential context for it. | {"state":{"query":"How does HyperMem train its JEV reasoning layer?","candidates":[{"index":0,"memory":"The DGUI_HYPERMEM-JEV dataset is the training brain for DGUI-HyperMem and grows with live usage."},{"index":1,"memory":"The JEV layer in DGUI-HyperMem records three kinds of training rows: analyze, rerank, and supers... | {"c0":{"noul":0},"c1":{"noul":0},"c2":{"noul":0.8},"c3":{"noul":0}} | {
"query": "How does HyperMem train its JEV reasoning layer?",
"candidates": [
{
"index": 0,
"memory": "The DGUI_HYPERMEM-JEV dataset is the training brain for DGUI-HyperMem and grows with live usage."
},
{
"index": 1,
"memory": "The JEV layer in DGUI-HyperMem records three kinds... | {
"c0": {
"type": "noul",
"instructions": "Memory candidate 0 is printed below under candidates[0]. Does it directly provide the information the query asks for, or essential context needed to act on it?",
"criteria": {
"true": "The candidate answers the query, or is necessary context for it, even if... | {
"c0": {
"noul": 0
},
"c1": {
"noul": 0
},
"c2": {
"noul": 0.8
},
"c3": {
"noul": 0
}
} | workers-ai | @cf/meta/llama-3.1-8b-fast-v2 | dgui-hypermem | null | null | 1,789,797,354,330 |
72e66bac-79cb-489d-b2cd-a0c25b07a4fe | analyze | choice_noul_score | Given the state (a memory to store), define and answer the TypeSafe System One questions that classify it: pick its type (choice), judge whether it is durable (noul), and score its salience (score). | {"state":{"memory":"Task completed: Technical whitepaper for DGUI-HyperMem written using arxiv-style LaTeX template with dual-column format. Title image from image0.png. Author: Wan Mohd Azizi Bin Wan Hosen (CTAXNAGOMI). Credits to TypeSafe AI (Jev/System One), DeckerGUI, KrackedDevs (krackeddevs.com), CTECX. Includes ... | {"memory_type":{"choice":"project","probabilities":{"fact":0.01,"decision":0,"architecture":0,"conversation":0,"project":0.96,"solution":0,"preference":0.03}},"durable":{"noul":0.92},"salience":{"score":2.83}} | {
"memory": "Task completed: Technical whitepaper for DGUI-HyperMem written using arxiv-style LaTeX template with dual-column format. Title image from image0.png. Author: Wan Mohd Azizi Bin Wan Hosen (CTAXNAGOMI). Credits to TypeSafe AI (Jev/System One), DeckerGUI, KrackedDevs (krackeddevs.com), CTECX. Includes keywo... | {
"memory_type": {
"type": "choice",
"instructions": "Which single category best describes this memory?",
"criteria": {
"fact": "A durable statement of fact about the user, their environment, tools, or the world.",
"preference": "A stated like, dislike, style, or way of working.",
"decis... | {
"memory_type": {
"choice": "project",
"probabilities": {
"fact": 0.01,
"decision": 0,
"architecture": 0,
"conversation": 0,
"project": 0.96,
"solution": 0,
"preference": 0.03
}
},
"durable": {
"noul": 0.92
},
"salience": {
"score": 2.83
}
} | typesafe | jev-latest | default | 008b1fa6-1143-4bb3-809d-2371d1ee3937 | opencode-dgui-hypermem-deploy | 1,789,890,334,227 |
29a2e58a-c399-48a7-96c4-bec8df7fb9dc | analyze | choice_noul_score | Given the state (a memory to store), define and answer the TypeSafe System One questions that classify it: pick its type (choice), judge whether it is durable (noul), and score its salience (score). | {"state":{"memory":"DGUI-HyperMem whitepaper successfully compiled to PDF at D:\\dgui-cli\\draft\\dguihypermem-whitepaper\\whitepaper.pdf. 5 pages, 1.8MB. Compilation: pdflatex (MiKTeX on Windows). Used arxiv-style template with twocolumn option. For SVG-to-PNG conversion: use Python Pillow to generate badge images whe... | {"memory_type":{"choice":"project","probabilities":{"solution":0.01,"architecture":0,"decision":0,"preference":0,"fact":0.02,"conversation":0,"project":0.97}},"durable":{"noul":0.78},"salience":{"score":2.14}} | {
"memory": "DGUI-HyperMem whitepaper successfully compiled to PDF at D:\\dgui-cli\\draft\\dguihypermem-whitepaper\\whitepaper.pdf. 5 pages, 1.8MB. Compilation: pdflatex (MiKTeX on Windows). Used arxiv-style template with twocolumn option. For SVG-to-PNG conversion: use Python Pillow to generate badge images when cai... | {
"memory_type": {
"type": "choice",
"instructions": "Which single category best describes this memory?",
"criteria": {
"fact": "A durable statement of fact about the user, their environment, tools, or the world.",
"preference": "A stated like, dislike, style, or way of working.",
"decis... | {
"memory_type": {
"choice": "project",
"probabilities": {
"solution": 0.01,
"architecture": 0,
"decision": 0,
"preference": 0,
"fact": 0.02,
"conversation": 0,
"project": 0.97
}
},
"durable": {
"noul": 0.78
},
"salience": {
"score": 2.14
}
} | typesafe | jev-latest | default | 67b733d8-0155-4c9e-afa1-ada57b72c526 | opencode-dgui-hypermem-deploy | 1,789,892,930,754 |
0dc7ca35-ff23-4324-add7-b4c3e1105469 | analyze | choice_noul_score | Given the state (a memory to store), define and answer the TypeSafe System One questions that classify it: pick its type (choice), judge whether it is durable (noul), and score its salience (score). | {"state":{"memory":"DGUI-HyperMem whitepaper v2 successfully compiled to PDF. 5 pages, dual-column arxiv-style format. Key lessons for future: (1) DO NOT use \\documentclass[twocolumn] WITH \\usepackage{arxiv} - they conflict on geometry. Instead use standalone twocolumn with manual fancyhdr setup. (2) For images: use ... | {"memory_type":{"choice":"project","probabilities":{"architecture":0,"solution":0.38,"preference":0,"conversation":0,"project":0.61,"decision":0.01,"fact":0}},"durable":{"noul":0.89},"salience":{"score":2.51}} | {
"memory": "DGUI-HyperMem whitepaper v2 successfully compiled to PDF. 5 pages, dual-column arxiv-style format. Key lessons for future: (1) DO NOT use \\documentclass[twocolumn] WITH \\usepackage{arxiv} - they conflict on geometry. Instead use standalone twocolumn with manual fancyhdr setup. (2) For images: use \\beg... | {
"memory_type": {
"type": "choice",
"instructions": "Which single category best describes this memory?",
"criteria": {
"fact": "A durable statement of fact about the user, their environment, tools, or the world.",
"preference": "A stated like, dislike, style, or way of working.",
"decis... | {
"memory_type": {
"choice": "project",
"probabilities": {
"architecture": 0,
"solution": 0.38,
"preference": 0,
"conversation": 0,
"project": 0.61,
"decision": 0.01,
"fact": 0
}
},
"durable": {
"noul": 0.89
},
"salience": {
"score": 2.51
}
} | typesafe | jev-latest | default | a0a923e2-6b46-4a41-8ce5-555bfc898385 | opencode-dgui-hypermem-deploy | 1,789,894,052,724 |
32267e9a-12df-4857-ab34-75ce286cf757 | analyze | choice_noul_score | Given the state (a memory to store), define and answer the TypeSafe System One questions that classify it: pick its type (choice), judge whether it is durable (noul), and score its salience (score). | {"state":{"memory":"DGUI-HyperMem whitepaper v3 compiled successfully: removed full-width title image, added JEV benchmark section with real data from TypeSafe docs (CLERC re-ranking: 5% to 18% top-1, parallel questions: 12.2x cheaper, cost comparison: 193.6x faster/444.6x cheaper vs LLMs). Used patterns: Speculative F... | {"memory_type":{"choice":"project","probabilities":{"project":0.63,"fact":0.23,"preference":0,"architecture":0.04,"solution":0.05,"conversation":0.03,"decision":0.02}},"durable":{"noul":0.8},"salience":{"score":2.8}} | {
"memory": "DGUI-HyperMem whitepaper v3 compiled successfully: removed full-width title image, added JEV benchmark section with real data from TypeSafe docs (CLERC re-ranking: 5% to 18% top-1, parallel questions: 12.2x cheaper, cost comparison: 193.6x faster/444.6x cheaper vs LLMs). Used patterns: Speculative Fan-Ou... | {
"memory_type": {
"type": "choice",
"instructions": "Which single category best describes this memory?",
"criteria": {
"fact": "A durable statement of fact about the user, their environment, tools, or the world.",
"preference": "A stated like, dislike, style, or way of working.",
"decis... | {
"memory_type": {
"choice": "project",
"probabilities": {
"project": 0.63,
"fact": 0.23,
"preference": 0,
"architecture": 0.04,
"solution": 0.05,
"conversation": 0.03,
"decision": 0.02
}
},
"durable": {
"noul": 0.8
},
"salience": {
"score": 2.8
... | typesafe | jev-latest | default | ee016d2e-eb3f-412f-a794-841351bcab84 | opencode-dgui-hypermem-deploy | 1,789,895,215,375 |
8a63b4ec-d30a-4824-9387-f5bb46be8241 | analyze | choice_noul_score | Given the state (a memory to store), define and answer the TypeSafe System One questions that classify it: pick its type (choice), judge whether it is durable (noul), and score its salience (score). | {"state":{"memory":"Error 1014 CNAME Cross-User Banned: Cannot CNAME from deckergui.my (wan.mohd.azizi.seggaf Cloudflare account) to dguihymem.ctaxnagomi.workers.dev (ctaxnagomi Cloudflare account). Cloudflare blocks CNAME across accounts. Fix: deploy all Workers (dgui-hypermem AND dguihymem) under the SAME Cloudflare ... | {"memory_type":{"choice":"solution","probabilities":{"decision":0,"preference":0,"solution":1,"architecture":0,"fact":0,"project":0,"conversation":0}},"durable":{"noul":0.92},"salience":{"score":3.49}} | {
"memory": "Error 1014 CNAME Cross-User Banned: Cannot CNAME from deckergui.my (wan.mohd.azizi.seggaf Cloudflare account) to dguihymem.ctaxnagomi.workers.dev (ctaxnagomi Cloudflare account). Cloudflare blocks CNAME across accounts. Fix: deploy all Workers (dgui-hypermem AND dguihymem) under the SAME Cloudflare accou... | {
"memory_type": {
"type": "choice",
"instructions": "Which single category best describes this memory?",
"criteria": {
"fact": "A durable statement of fact about the user, their environment, tools, or the world.",
"preference": "A stated like, dislike, style, or way of working.",
"decis... | {
"memory_type": {
"choice": "solution",
"probabilities": {
"decision": 0,
"preference": 0,
"solution": 1,
"architecture": 0,
"fact": 0,
"project": 0,
"conversation": 0
}
},
"durable": {
"noul": 0.92
},
"salience": {
"score": 3.49
}
} | typesafe | jev-latest | default | a0a5eba6-0b7d-461a-9407-0282c72fe5a0 | opencode-dgui-hypermem-deploy | 1,789,897,231,681 |
337363d9-07be-47ec-8c53-a2e51e6c9a49 | analyze | choice_noul_score | Given the state (a memory to store), define and answer the TypeSafe System One questions that classify it: pick its type (choice), judge whether it is durable (noul), and score its salience (score). | {"state":{"memory":"DGUI-HyperMem migration to wan.mohd.azizi.seggaf Cloudflare account (ID: 155c4982c49d57ce2ff8c5a27e599cbd). D1 database UUID: fb61c1d7-ded8-485f-9d3b-2b6fddf9b979 (database name: dgui-hypermem). Vectorize index NOT YET created on wan account. CRM token with full permissions (168 resources) created b... | {"memory_type":{"choice":"project","probabilities":{"project":1,"decision":0,"solution":0,"architecture":0,"conversation":0,"preference":0,"fact":0}},"durable":{"noul":0.84},"salience":{"score":3.55}} | {
"memory": "DGUI-HyperMem migration to wan.mohd.azizi.seggaf Cloudflare account (ID: 155c4982c49d57ce2ff8c5a27e599cbd). D1 database UUID: fb61c1d7-ded8-485f-9d3b-2b6fddf9b979 (database name: dgui-hypermem). Vectorize index NOT YET created on wan account. CRM token with full permissions (168 resources) created but no... | {
"memory_type": {
"type": "choice",
"instructions": "Which single category best describes this memory?",
"criteria": {
"fact": "A durable statement of fact about the user, their environment, tools, or the world.",
"preference": "A stated like, dislike, style, or way of working.",
"decis... | {
"memory_type": {
"choice": "project",
"probabilities": {
"project": 1,
"decision": 0,
"solution": 0,
"architecture": 0,
"conversation": 0,
"preference": 0,
"fact": 0
}
},
"durable": {
"noul": 0.84
},
"salience": {
"score": 3.55
}
} | typesafe | jev-latest | default | 0e857784-f515-4bb6-8203-ddcea8ba4676 | opencode-dgui-hypermem-deploy | 1,789,901,044,759 |
DGUI_HYPERMEM-JEV
The training "brain" for DGUI-HyperMem (DeckerGUI HyperMemory) — the self-hosted memory MCP server. Every JEV reasoning decision the service makes is appended here as a typed instruction row, so the corpus grows with real usage and can be used to fine-tune or few-shot the JEV layer later.
Usage
from datasets import load_dataset
ds = load_dataset("ctaxnagomi/DGUI_HYPERMEM-JEV", split="train")
for row in ds.stream():
print(row["use_case"], row["instruct_type"], row["provider"])
Row types
| use_case | instruct_type | What was recorded |
|---|---|---|
| analyze | choice_noul_score | Classifying a new memory (type, durability, salience) |
| rerank | noul | Re-ranking query candidates during recall |
| supersede | noul | Contradiction checks that supersede stale memories |
Each row carries state, questions (the exact TypeSafe System One block that was sent),
answers (what came back), plus provider, model, scope, memory_id, source,
instruction, input, output, and created_at.
Credits
DGUI-HyperMem is a DeckerGUI project.
| Who | Contribution | Link |
|---|---|---|
| TypeSafe AI | Jev / System One, the Choice / Noul / Score primitives the JEV layer is built on | https://typesafe.ai |
| DeckerGUI | Design, implementation and operation | https://deckergui.my |
| KrackedDevs | Community credit and support | |
| CTECX | Knowledge / corpus partner |
Jev, System One, and the Choice / Noul / Score primitives are TypeSafe AI's and are used under their MIT-licensed public documentation at https://docs.typesafe.ai.
- Downloads last month
- 54