memworthy 0.1.0__tar.gz

memworthy-0.1.0/.gitignore

@@ -0,0 +1,36 @@
+# Private planning docs stay local; decisions.md is part of the repo.
+/docs/*
+!/docs/decisions.md
+
+# Secrets
+.env
+.env.*
+!.env.example
+
+# Python
+.venv/
+__pycache__/
+*.py[cod]
+*.egg-info/
+build/
+dist/
+.pytest_cache/
+.mypy_cache/
+.ruff_cache/
+.coverage
+htmlcov/
+
+# Runtime output
+memworthy.ledger.jsonl
+*.ledger.jsonl
+!tests/fixtures/**/*.ledger.jsonl
+playground/backend/.state/
+node_modules/
+.DS_Store
+.remember/
+
+
+# LongMemEval run artifacts (cache, per-question rows); only README.md is tracked
+/eval/results/longmemeval/*
+!/eval/results/longmemeval/README.md
+*.tsbuildinfo

memworthy-0.1.0/LICENSE

@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

memworthy-0.1.0/PKG-INFO

@@ -0,0 +1,228 @@
+Metadata-Version: 2.5
+Name: memworthy
+Version: 0.1.0
+Summary: Decide what an AI system should remember: policy-driven memory gating with typed judgments.
+Project-URL: Homepage, https://github.com/AfnanHussain10/Memworthy
+Project-URL: Repository, https://github.com/AfnanHussain10/Memworthy
+Project-URL: Issues, https://github.com/AfnanHussain10/Memworthy/issues
+Author: Afnan Hussain
+License-Expression: Apache-2.0
+License-File: LICENSE
+Keywords: agents,jev,llm,memory,policy
+Classifier: License :: OSI Approved :: Apache Software License
+Classifier: Programming Language :: Python :: 3
+Classifier: Typing :: Typed
+Requires-Python: >=3.10
+Requires-Dist: httpx>=0.27
+Requires-Dist: pydantic>=2.6
+Requires-Dist: pyyaml>=6
+Requires-Dist: typer>=0.12
+Provides-Extra: dev
+Requires-Dist: fastapi>=0.110; extra == 'dev'
+Requires-Dist: matplotlib>=3.8; extra == 'dev'
+Requires-Dist: mypy>=1.10; extra == 'dev'
+Requires-Dist: pytest-asyncio>=0.23; extra == 'dev'
+Requires-Dist: pytest-cov>=5; extra == 'dev'
+Requires-Dist: pytest>=8; extra == 'dev'
+Requires-Dist: python-dotenv>=1; extra == 'dev'
+Requires-Dist: ruff>=0.6; extra == 'dev'
+Requires-Dist: types-pyyaml; extra == 'dev'
+Requires-Dist: uvicorn>=0.29; extra == 'dev'
+Provides-Extra: jev
+Provides-Extra: mem0
+Requires-Dist: mem0ai<2.3,>=2.2; extra == 'mem0'
+Description-Content-Type: text/markdown
+
+# Memworthy
+
+**Memory should be decided, not just extracted.**
+
+Agent memory usually fails at the decision step, not at retrieval. Extraction pipelines are generous, so memory fills up with errors:
+- temporary states stored as permanent facts ("in Dubai this week" becomes "lives in Dubai")
+- plans stored as facts
+- other people's facts stored as the user's
+- stale values and ignored corrections
+- the assistant's guesses saved as things the user said
+- secrets
+- no record of why anything was kept
+
+Memworthy is a small Python library that sits between whatever proposes memories and whatever stores them. For each candidate it:
+1. runs deterministic code checks
+2. asks TypeSafe's Jev model one batch of typed questions
+3. applies an ordered, user-editable policy to choose an action: `store`, `update`, `merge`, `supersede`, `reject`, `review`, `redact` or `store_labeled`
+
+Every decision is written to a ledger you can audit, replay under a changed policy, and review.
+
+## Quickstart
+
+```bash
+pip install "memworthy[jev]"              # Python 3.10+
+export TYPESAFE_API_KEY=...             # https://typesafe.ai
+```
+
+```python
+from memworthy import Candidate, DictStore, Gate
+
+gate = Gate(policy="personal-memory", store=DictStore())
+decisions = gate.ingest_messages([
+    {"role": "user", "content": "I live in Dubai"},
+    {"role": "user", "content": "Flying to Riyadh tomorrow"},
+    {"role": "user", "content": "I moved to Riyadh last week"},
+    {"role": "user", "content": "Actually I'm only considering it"},
+])
+for d in decisions:
+    print(f"{d.action:10} {d.rule:16} {d.candidate.text}")
+# store      rules[9]         I live in Dubai
+# reject     not_memory       Flying to Riyadh tomorrow
+# update     update_existing  I moved to Riyadh last week
+# supersede  retracted        Actually I'm only considering it   (rolls back to Dubai)
+```
+
+The output above is what the bundled recorded Jev answers give; a live run may differ slightly. You can reproduce it with no API key:
+
+```python
+from memworthy import RecordedJudge
+from memworthy.judges.recorded import bundled_fixture_path
+
+gate = Gate("personal-memory", DictStore(),
+            judge=RecordedJudge(bundled_fixture_path("personal-memory")))
+```
+
+Each decision also lands in `memworthy.ledger.jsonl` with every signal, its probability, the rule that fired, and the policy and model version.
+
+### Other integrations
+
+```python
+# Wrap a Mem0 client: only approved content reaches Mem0 (pip install "memworthy[jev,mem0]")
+from mem0 import Memory
+from memworthy import GatedMemory
+memory = GatedMemory(Memory(), policy="personal-memory")          # mode="input" or "candidate"
+memory.add([{"role": "user", "content": "I'm in Lisbon this week"}], user_id="alice")  # rejected
+```
+
+```bash
+# Turn Claude Code / Codex sessions into an Obsidian-compatible Markdown knowledge base
+memworthy run dev-sessions ~/.claude/projects --store markdown:~/kb
+memworthy run dev-sessions ~/.claude/projects --store markdown:~/kb --extractor llm   # OpenRouter via EXTRACTOR_MODEL
+```
+
+## CLI
+
+| Command | What it does |
+| --- | --- |
+| `memworthy test POLICY [--judge recorded\|jev\|mock] [--pairs FILE]` | Runs the policy's tests; exits 1 on failure (CI-friendly with recorded answers) |
+| `memworthy run POLICY SOURCE [--store dict\|markdown:PATH\|mem0] [--dry-run]` | Gates a candidate file, chat JSON or session folder |
+| `memworthy replay LEDGER POLICY [--only-changed]` | Shows which past decisions a policy change would flip (no API calls when prompts are unchanged) |
+| `memworthy lint POLICY [--strict]` | Overlapping types, unused signals, unreachable rules, vague prompts |
+| `memworthy review LEDGER [--export-tests FILE]` | Accept, override or skip queued decisions; overrides become tests |
+| `memworthy record POLICY INPUT` | Records live Jev answers as fixtures |
+| `memworthy eval POLICY` | Contrast-pair metrics, calibration and plots into `eval/results/` |
+
+## Templates
+
+Both are ordinary YAML files in `src/memworthy/templates/`; copy one and edit it.
+
+**`personal-memory`** (v1.1) is for consumer assistants.
+- **Types:** fact, preference, relationship, temporary, intent, hypothetical, noise.
+- **Rules, in order:**
+  1. Redact secrets.
+  2. Reject anything the user didn't say.
+  3. Roll back retractions.
+  4. Reject temporary states, plans, hypotheticals and noise.
+  5. Reject future or time-bounded statements. Dates are parsed by code, not the judge.
+  6. Review sensitive data (health, finances, religion, precise location).
+  7. Reject low-durability facts.
+  8. Review facts about someone else.
+  9. Update memories the new fact replaces.
+  10. Store the rest.
+
+```yaml
+rules:
+  - {name: secrets, when: "secret", then: redact}
+  - {name: not_user, when: "not from_user", then: reject}
+  - {name: retracted, when: "retraction > 0.7 and conflict", then: supersede, restore: true}
+  - {name: not_memory, when: "type in ['temporary', 'intent', 'hypothetical', 'noise']", then: reject}
+  - {name: bounded_time, when: "timing == 'bounded' or timing == 'future'", then: reject}
+  - {name: sensitive, when: "sensitivity >= 1.5", then: review}
+  - {name: low_durability, when: "durable < 0.6 or explicit < 0.5", then: reject}
+  - {name: wrong_subject, when: "about_subject < 0.5", then: review}
+  - {name: update_existing, when: "conflict and conflict.p > 0.6", then: update}
+  - {else: store}
+```
+
+**`dev-sessions`** (v1.1) turns coding-agent sessions into a knowledge base.
+- **Pipeline:**
+  1. Sessions are split into episodes.
+  2. Each episode is classified with one Jev call, and noise is dropped.
+  3. Candidates are extracted.
+  4. Each candidate is gated.
+- **Types:** task, fix, decision, discussion, explanation, concept, convention, noise.
+- **What happens to each kind:**
+  - Unverified fixes and tasks are labeled `unverified`, meaning no passing tests, build, commit or user sign-off.
+  - Explanations not grounded in files read are labeled.
+  - Unfinished discussions become open questions.
+  - Proposals the user did not endorse go to review.
+  - Abandoned approaches and generic textbook content are rejected.
+  - Later decisions supersede earlier ones, and the old entry moves to `_archive/`.
+  - Secrets are redacted before the judge sees them.
+
+Policies use a small, safe expression language (a hand-written parser, never `eval`). For anything it can't express, register Python checks and rules with `@memworthy.check` and `@memworthy.rule`.
+
+## Evaluation results
+
+All numbers come from real `jev-1.13.0` answers, regenerated with `memworthy eval <policy>`. Full reports, including every failure, are in [`eval/results/`](https://github.com/AfnanHussain10/Memworthy/blob/main/eval/results/README.md).
+
+Contrast pairs are held out: one changed word flips the correct action, labels come from the template, and the policies were frozen before the pairs were recorded.
+
+| Template | Cases | Action accuracy | Pair consistency | Target |
+| --- | --- | --- | --- | --- |
+| `personal-memory` v1.1 | 272 | 97.4% | 94.4% | ≥90% / ≥85%: met |
+| `dev-sessions` v1.1 | 236 | 91.5% | 83.9% | pair consistency misses 85% |
+
+- **First run.** Three pair templates had labeling or wording errors and were corrected after the first run. That run scored 94.5% / 88.1% and 87.3% / 75.4%; it is kept in the repo, and the corrections are listed in `eval/results/README.md`.
+- **Remaining failures (genuine):**
+  - A question that embeds a fact ("Since I live in Vienna, what…") is read as an intent.
+  - Failed fixes are rejected as noise instead of being labeled unverified.
+  - Some project-specific facts are typed as explanations and labeled ungrounded.
+- **Latency and cost.** The median Jev round trip was about 1 s from the author's machine, which misses the 500 ms target. The median cost is about $0.00003 per candidate.
+- **Calibration.** Raw ECE is 0.05–0.29 on small samples. See the reports for what the temperature-scaled numbers do and do not mean.
+- **Built-in tests are not evidence.** The template tests (124 and 105) pass 100%, but they were used to tune the prompts.
+- **LongMemEval knowledge-update, with vs without Memworthy: pending real run.** Run it with `python eval/longmemeval.py longmemeval_s_cleaned.json --limit 78`; details are in `eval/results/longmemeval/README.md`.
+
+## Playground
+
+`playground/` has a FastAPI backend that imports this library and a React front end:
+- a guided Dubai → Riyadh demo, one timeline row per message: message → verdict → effect on memory, with "Why?" showing the rule, judge answers and thresholds
+- free play with example messages
+- a policy panel with one-click experiments, rule switches and a YAML editor; edits replay the conversation and mark changed decisions
+- the recorded coding-session demo
+- the evaluation page
+
+```bash
+pip install -e ".[dev,jev]"
+uvicorn playground.backend.app:app --port 8765     # open http://localhost:8765
+```
+
+The built front end is committed in `playground/web/`. To change it, edit `playground/ui/` and rebuild:
+
+```bash
+cd playground/ui && npm ci
+npm run dev        # http://localhost:5173, proxies /api to the backend on :8765
+npm run build      # typechecks, then writes ../web
+```
+
+Replay mode uses the bundled recorded answers and needs no key. Live mode needs `TYPESAFE_API_KEY`. It is rate-limited per IP (40 calls per hour) and globally (600 per hour), and has a daily spend cap (`PLAYGROUND_DAILY_CAP_USD`, default $0.50). Every decision is labeled recorded, live or needs-live.
+
+## Development
+
+```bash
+pyenv install -s 3.12 && python -m venv .venv
+.venv/bin/pip install -e ".[dev,jev,mem0]"
+.venv/bin/pytest -q && .venv/bin/mypy src && .venv/bin/ruff check .
+```
+
+Tests never call live APIs; `pytest -m live` runs a few real Jev calls when `TYPESAFE_API_KEY` is set. Design decisions are in `docs/decisions.md` and build status is in `PROGRESS.md`.
+
+## License
+
+Apache 2.0.

memworthy-0.1.0/README.md

@@ -0,0 +1,193 @@
+# Memworthy
+
+**Memory should be decided, not just extracted.**
+
+Agent memory usually fails at the decision step, not at retrieval. Extraction pipelines are generous, so memory fills up with errors:
+- temporary states stored as permanent facts ("in Dubai this week" becomes "lives in Dubai")
+- plans stored as facts
+- other people's facts stored as the user's
+- stale values and ignored corrections
+- the assistant's guesses saved as things the user said
+- secrets
+- no record of why anything was kept
+
+Memworthy is a small Python library that sits between whatever proposes memories and whatever stores them. For each candidate it:
+1. runs deterministic code checks
+2. asks TypeSafe's Jev model one batch of typed questions
+3. applies an ordered, user-editable policy to choose an action: `store`, `update`, `merge`, `supersede`, `reject`, `review`, `redact` or `store_labeled`
+
+Every decision is written to a ledger you can audit, replay under a changed policy, and review.
+
+## Quickstart
+
+```bash
+pip install "memworthy[jev]"              # Python 3.10+
+export TYPESAFE_API_KEY=...             # https://typesafe.ai
+```
+
+```python
+from memworthy import Candidate, DictStore, Gate
+
+gate = Gate(policy="personal-memory", store=DictStore())
+decisions = gate.ingest_messages([
+    {"role": "user", "content": "I live in Dubai"},
+    {"role": "user", "content": "Flying to Riyadh tomorrow"},
+    {"role": "user", "content": "I moved to Riyadh last week"},
+    {"role": "user", "content": "Actually I'm only considering it"},
+])
+for d in decisions:
+    print(f"{d.action:10} {d.rule:16} {d.candidate.text}")
+# store      rules[9]         I live in Dubai
+# reject     not_memory       Flying to Riyadh tomorrow
+# update     update_existing  I moved to Riyadh last week
+# supersede  retracted        Actually I'm only considering it   (rolls back to Dubai)
+```
+
+The output above is what the bundled recorded Jev answers give; a live run may differ slightly. You can reproduce it with no API key:
+
+```python
+from memworthy import RecordedJudge
+from memworthy.judges.recorded import bundled_fixture_path
+
+gate = Gate("personal-memory", DictStore(),
+            judge=RecordedJudge(bundled_fixture_path("personal-memory")))
+```
+
+Each decision also lands in `memworthy.ledger.jsonl` with every signal, its probability, the rule that fired, and the policy and model version.
+
+### Other integrations
+
+```python
+# Wrap a Mem0 client: only approved content reaches Mem0 (pip install "memworthy[jev,mem0]")
+from mem0 import Memory
+from memworthy import GatedMemory
+memory = GatedMemory(Memory(), policy="personal-memory")          # mode="input" or "candidate"
+memory.add([{"role": "user", "content": "I'm in Lisbon this week"}], user_id="alice")  # rejected
+```
+
+```bash
+# Turn Claude Code / Codex sessions into an Obsidian-compatible Markdown knowledge base
+memworthy run dev-sessions ~/.claude/projects --store markdown:~/kb
+memworthy run dev-sessions ~/.claude/projects --store markdown:~/kb --extractor llm   # OpenRouter via EXTRACTOR_MODEL
+```
+
+## CLI
+
+| Command | What it does |
+| --- | --- |
+| `memworthy test POLICY [--judge recorded\|jev\|mock] [--pairs FILE]` | Runs the policy's tests; exits 1 on failure (CI-friendly with recorded answers) |
+| `memworthy run POLICY SOURCE [--store dict\|markdown:PATH\|mem0] [--dry-run]` | Gates a candidate file, chat JSON or session folder |
+| `memworthy replay LEDGER POLICY [--only-changed]` | Shows which past decisions a policy change would flip (no API calls when prompts are unchanged) |
+| `memworthy lint POLICY [--strict]` | Overlapping types, unused signals, unreachable rules, vague prompts |
+| `memworthy review LEDGER [--export-tests FILE]` | Accept, override or skip queued decisions; overrides become tests |
+| `memworthy record POLICY INPUT` | Records live Jev answers as fixtures |
+| `memworthy eval POLICY` | Contrast-pair metrics, calibration and plots into `eval/results/` |
+
+## Templates
+
+Both are ordinary YAML files in `src/memworthy/templates/`; copy one and edit it.
+
+**`personal-memory`** (v1.1) is for consumer assistants.
+- **Types:** fact, preference, relationship, temporary, intent, hypothetical, noise.
+- **Rules, in order:**
+  1. Redact secrets.
+  2. Reject anything the user didn't say.
+  3. Roll back retractions.
+  4. Reject temporary states, plans, hypotheticals and noise.
+  5. Reject future or time-bounded statements. Dates are parsed by code, not the judge.
+  6. Review sensitive data (health, finances, religion, precise location).
+  7. Reject low-durability facts.
+  8. Review facts about someone else.
+  9. Update memories the new fact replaces.
+  10. Store the rest.
+
+```yaml
+rules:
+  - {name: secrets, when: "secret", then: redact}
+  - {name: not_user, when: "not from_user", then: reject}
+  - {name: retracted, when: "retraction > 0.7 and conflict", then: supersede, restore: true}
+  - {name: not_memory, when: "type in ['temporary', 'intent', 'hypothetical', 'noise']", then: reject}
+  - {name: bounded_time, when: "timing == 'bounded' or timing == 'future'", then: reject}
+  - {name: sensitive, when: "sensitivity >= 1.5", then: review}
+  - {name: low_durability, when: "durable < 0.6 or explicit < 0.5", then: reject}
+  - {name: wrong_subject, when: "about_subject < 0.5", then: review}
+  - {name: update_existing, when: "conflict and conflict.p > 0.6", then: update}
+  - {else: store}
+```
+
+**`dev-sessions`** (v1.1) turns coding-agent sessions into a knowledge base.
+- **Pipeline:**
+  1. Sessions are split into episodes.
+  2. Each episode is classified with one Jev call, and noise is dropped.
+  3. Candidates are extracted.
+  4. Each candidate is gated.
+- **Types:** task, fix, decision, discussion, explanation, concept, convention, noise.
+- **What happens to each kind:**
+  - Unverified fixes and tasks are labeled `unverified`, meaning no passing tests, build, commit or user sign-off.
+  - Explanations not grounded in files read are labeled.
+  - Unfinished discussions become open questions.
+  - Proposals the user did not endorse go to review.
+  - Abandoned approaches and generic textbook content are rejected.
+  - Later decisions supersede earlier ones, and the old entry moves to `_archive/`.
+  - Secrets are redacted before the judge sees them.
+
+Policies use a small, safe expression language (a hand-written parser, never `eval`). For anything it can't express, register Python checks and rules with `@memworthy.check` and `@memworthy.rule`.
+
+## Evaluation results
+
+All numbers come from real `jev-1.13.0` answers, regenerated with `memworthy eval <policy>`. Full reports, including every failure, are in [`eval/results/`](https://github.com/AfnanHussain10/Memworthy/blob/main/eval/results/README.md).
+
+Contrast pairs are held out: one changed word flips the correct action, labels come from the template, and the policies were frozen before the pairs were recorded.
+
+| Template | Cases | Action accuracy | Pair consistency | Target |
+| --- | --- | --- | --- | --- |
+| `personal-memory` v1.1 | 272 | 97.4% | 94.4% | ≥90% / ≥85%: met |
+| `dev-sessions` v1.1 | 236 | 91.5% | 83.9% | pair consistency misses 85% |
+
+- **First run.** Three pair templates had labeling or wording errors and were corrected after the first run. That run scored 94.5% / 88.1% and 87.3% / 75.4%; it is kept in the repo, and the corrections are listed in `eval/results/README.md`.
+- **Remaining failures (genuine):**
+  - A question that embeds a fact ("Since I live in Vienna, what…") is read as an intent.
+  - Failed fixes are rejected as noise instead of being labeled unverified.
+  - Some project-specific facts are typed as explanations and labeled ungrounded.
+- **Latency and cost.** The median Jev round trip was about 1 s from the author's machine, which misses the 500 ms target. The median cost is about $0.00003 per candidate.
+- **Calibration.** Raw ECE is 0.05–0.29 on small samples. See the reports for what the temperature-scaled numbers do and do not mean.
+- **Built-in tests are not evidence.** The template tests (124 and 105) pass 100%, but they were used to tune the prompts.
+- **LongMemEval knowledge-update, with vs without Memworthy: pending real run.** Run it with `python eval/longmemeval.py longmemeval_s_cleaned.json --limit 78`; details are in `eval/results/longmemeval/README.md`.
+
+## Playground
+
+`playground/` has a FastAPI backend that imports this library and a React front end:
+- a guided Dubai → Riyadh demo, one timeline row per message: message → verdict → effect on memory, with "Why?" showing the rule, judge answers and thresholds
+- free play with example messages
+- a policy panel with one-click experiments, rule switches and a YAML editor; edits replay the conversation and mark changed decisions
+- the recorded coding-session demo
+- the evaluation page
+
+```bash
+pip install -e ".[dev,jev]"
+uvicorn playground.backend.app:app --port 8765     # open http://localhost:8765
+```
+
+The built front end is committed in `playground/web/`. To change it, edit `playground/ui/` and rebuild:
+
+```bash
+cd playground/ui && npm ci
+npm run dev        # http://localhost:5173, proxies /api to the backend on :8765
+npm run build      # typechecks, then writes ../web
+```
+
+Replay mode uses the bundled recorded answers and needs no key. Live mode needs `TYPESAFE_API_KEY`. It is rate-limited per IP (40 calls per hour) and globally (600 per hour), and has a daily spend cap (`PLAYGROUND_DAILY_CAP_USD`, default $0.50). Every decision is labeled recorded, live or needs-live.
+
+## Development
+
+```bash
+pyenv install -s 3.12 && python -m venv .venv
+.venv/bin/pip install -e ".[dev,jev,mem0]"
+.venv/bin/pytest -q && .venv/bin/mypy src && .venv/bin/ruff check .
+```
+
+Tests never call live APIs; `pytest -m live` runs a few real Jev calls when `TYPESAFE_API_KEY` is set. Design decisions are in `docs/decisions.md` and build status is in `PROGRESS.md`.
+
+## License
+
+Apache 2.0.