diff --git a/README.md b/README.md index 4cbaa24..98c3b44 100644 --- a/README.md +++ b/README.md @@ -87,10 +87,11 @@ What you get: - a domain pack for the topic - a guided curriculum path -- a synthetic learner run over that path +- a deterministic learner run over that path - a capability export - a reusable skill bundle - visual progress artifacts +- an optional local-LLM learner/mentor transcript path via RoleMesh This is the best "show me why this is fun" path in the current repo. @@ -103,9 +104,9 @@ If you already have notes, a syllabus, or a course outline, the lightest custom 3. Emit a draft pack. 4. Review only what matters. -The easiest reference for this flow is the OCW demo source: +The easiest reference for this flow is the OCW demo source tree: -- `examples/ocw-information-entropy/6-050j-information-and-entropy.md` +- `examples/ocw-information-entropy/course/` Use it as a template for your own topic, then follow the same pattern implemented in: diff --git a/docs/faq.md b/docs/faq.md index 9fac81b..4b005b1 100644 --- a/docs/faq.md +++ b/docs/faq.md @@ -98,7 +98,12 @@ Yes, but conservatively. Those adapters currently normalize text in a simplified ## Does the agentic learner call an external LLM? -No. The current agentic learner paths are deterministic and synthetic. They are meant to exercise the orchestration pattern, evaluator pipeline, mastery updates, capability export, and visualization flow without requiring an external model service. +There are now two learner paths in the repo. + +- The main pack/demo learner loop is still deterministic. It exercises orchestration, evaluator flow, mastery updates, capability export, and visualization without requiring an external model service. +- There is also a local-LLM learner path through the RoleMesh integration and transcript demo. That path is intended to show how mentor, learner, practice, and evaluator roles can run against a local model gateway. + +So the deterministic learner is still active, but it is no longer the only learner-style path shown in the repository. ## Can I still use it as a personal mentor even though the learner is synthetic? @@ -111,7 +116,7 @@ Right now the value is in: - exporting progress and capability artifacts - generating reusable skill context for future tutoring or evaluation -The current demos show the shape of a mentor workflow even though the agent itself is not yet a live external model integration. +The deterministic demos show the shape of a mentor workflow, and the RoleMesh transcript path shows the same pattern with a live local-LLM-backed learner role. ## How should I use it if I am taking a course and do not want to hire a tutor? diff --git a/skills/ocw-information-entropy-agent/SKILL.md b/skills/ocw-information-entropy-agent/SKILL.md index fb15dcd..b6c66dd 100644 --- a/skills/ocw-information-entropy-agent/SKILL.md +++ b/skills/ocw-information-entropy-agent/SKILL.md @@ -10,7 +10,7 @@ Use this skill when the task is about tutoring, evaluating, or planning study in ## Workflow 1. Read `references/generated-course-summary.md` for the pack structure and target concepts. -2. Read `references/generated-capability-summary.md` to understand what the demo AI learner already mastered. +2. Read `references/generated-capability-summary.md` to understand what the deterministic demo learner already mastered. 3. Use `assets/generated/pack/` as the source of truth for concept ids, prerequisites, and mastery signals. 4. Use `assets/generated/pack/source_corpus.json` to ground explanations in the ingested source material before relying on model prior knowledge. 5. When giving guidance, preserve the pack ordering from fundamentals through coding and thermodynamics. @@ -22,3 +22,4 @@ Use this skill when the task is about tutoring, evaluating, or planning study in - concept explanations tied to entropy, coding, and channel capacity - evaluation checklists using the generated capability report - follow-up exercises that extend the existing learner artifacts +- local-LLM tutoring or evaluation runs that use the same pack and source corpus through role-based prompts diff --git a/src/didactopus/ocw_information_entropy_demo.py b/src/didactopus/ocw_information_entropy_demo.py index e99605b..068cd2d 100644 --- a/src/didactopus/ocw_information_entropy_demo.py +++ b/src/didactopus/ocw_information_entropy_demo.py @@ -36,7 +36,7 @@ Use this skill when the task is about tutoring, evaluating, or planning study in ## Workflow 1. Read `references/generated-course-summary.md` for the pack structure and target concepts. -2. Read `references/generated-capability-summary.md` to understand what the demo AI learner already mastered. +2. Read `references/generated-capability-summary.md` to understand what the deterministic demo learner already mastered. 3. Use `assets/generated/pack/` as the source of truth for concept ids, prerequisites, and mastery signals. 4. Use `assets/generated/pack/source_corpus.json` to ground explanations in the ingested source material before relying on model prior knowledge. 5. When giving guidance, preserve the pack ordering from fundamentals through coding and thermodynamics. @@ -48,6 +48,7 @@ Use this skill when the task is about tutoring, evaluating, or planning study in - concept explanations tied to entropy, coding, and channel capacity - evaluation checklists using the generated capability report - follow-up exercises that extend the existing learner artifacts +- local-LLM tutoring or evaluation runs that use the same pack and source corpus through role-based prompts """