pmem-ai 0.6.2 → 0.7.0

package/skills/pmem/references/memory-cards.md

@@ -4,6 +4,8 @@ Memory cards are Markdown files with YAML frontmatter. They live under `.pmem/`
 
 ## Card Types
 
+The default `software` preset uses these common card types:
+
 | Type | Directory | Purpose |
 |------|-----------|---------|
 | `module` | `.pmem/modules/` | Code module description and ownership |
@@ -13,7 +15,9 @@ Memory cards are Markdown files with YAML frontmatter. They live under `.pmem/`
 | `risk` | `.pmem/risks/` | Identified risks and mitigations |
 | `trace` | `.pmem/traces/` | Work session traces |
 
-## Module Card (most important)
+v0.7.0 supports domain presets and custom schemas. A `novel` project can use `character`, `chapter`, `world`, and `arc`; a `research` project can use `source`, `claim`, `note`, and `experiment`. Check `.pmem/manifest.yml` → `schema.card_types`, `schema.type_dirs`, and `schema.creatable_types` for the project's valid types.
+
+## Software Module Card
 
 ```yaml
 ---
@@ -38,6 +42,45 @@ Handles user authentication and token management.
 
 The `source_files` field is critical — it links code files to memory. When `src/auth/index.ts` changes, pmem knows to flag `module.auth` as potentially stale.
 
+## Novel Character Card
+
+```yaml
+---
+id: character.protagonist
+type: character
+status: active
+tags: [main-cast]
+aliases: [hero]
+source_files: [draft/characters/protagonist.md]
+---
+# Protagonist
+
+## Role
+Primary viewpoint character.
+
+## Motivation
+Wants to recover a lost family archive.
+```
+
+## Research Source Card
+
+```yaml
+---
+id: source.smith_2024_survey
+type: source
+status: active
+tags: [memory, agents]
+source_files: [papers/smith-2024.pdf]
+---
+# Smith 2024 Survey
+
+## Summary
+Survey of persistent memory approaches for AI agents.
+
+## Evidence
+Supports claims about cross-session continuity.
+```
+
 ## Decision Card
 
 ```yaml

package/skills/pmem/references/session-workflow.md

@@ -14,13 +14,13 @@ pmem recall --format compact --budget 2000
 `pmem recall` output tells you:
 - Project name, stage, current focus
 - Recommended next step
-- Active modules to read
+- Active foundational cards to read (`active_foundation` in JSON)
 - Recent update history
 - Unresolved dirty flags count
 
 ## Before Focused Work
 
-Before working on a specific module or task:
+Before working on a specific module, character, source, task, or topic:
 
 ```bash
 pmem ask "<topic>" --format compact
@@ -51,7 +51,7 @@ Marks affected cards as potentially stale in the database.
 pmem update --suggest --format json
 ```
 
-Generates suggestions based on dirty flags and state freshness. **Exits with code 1 when suggestions exist** — this is a workflow signal, NOT a failure.
+Generates suggestions based on dirty flags and state freshness. v0.6.2+ exits with code `0` whether suggestions exist or not. Parse JSON fields such as `summary.has_actionable`, `summary.blocking`, or `summary.verify_blocking` instead of using exit code `1` as a workflow signal.
 
 ## Confirming Updates
 

package/skills/pmem/references/universal-domains.md

@@ -0,0 +1,66 @@
+# Universal Domains and Custom Schemas
+
+pmem v0.7.0 is domain-neutral. It still works for software codebases, but it can also initialize memory for writing, research, and custom agent projects.
+
+## Domain Presets
+
+```bash
+pmem init my-service --domain software
+pmem init my-novel --domain novel
+pmem init my-study --domain research
+```
+
+| Preset | Card directories | Foundational types | Discover |
+|--------|------------------|--------------------|----------|
+| `software` | `modules`, `features`, `decisions`, `tasks`, `risks`, `traces` | `module` | enabled |
+| `novel` | `characters`, `chapters`, `world`, `arc`, `decisions`, `traces` | `character`, `chapter` | disabled |
+| `research` | `sources`, `claims`, `notes`, `experiments`, `decisions`, `traces` | `source`, `claim` | disabled |
+
+## Manifest Schema
+
+The manifest controls valid card types and runtime behavior:
+
+```yaml
+schema:
+  card_types: [source, claim, note, experiment, decision, trace]
+  type_dirs:
+    source: sources
+    claim: claims
+    note: notes
+    experiment: experiments
+    decision: decisions
+    trace: traces
+  foundational_types: [source, claim]
+  evidence_types: [decision, trace]
+  creatable_types: [source, claim, note, experiment, decision, trace]
+  default_type: note
+discover:
+  enabled: false
+```
+
+Use `schema.creatable_types` to decide what `pmem new <type> <title>` may create. Use `schema.foundational_types` to decide what appears in recall as the project's core context.
+
+## Recall JSON
+
+For machine-readable recall:
+
+```bash
+pmem recall --format json
+```
+
+Read `active_foundation` for foundational cards. `active_modules` is kept as a compatibility alias and contains the same list.
+
+## Discovery
+
+`pmem discover` is useful for software projects because it scans imports and package dependencies. It is disabled by default for `novel` and `research` presets:
+
+```bash
+pmem discover --format json
+# exits 0 with a disabled message when discover.enabled is false
+```
+
+Enable it only when the project has source files that benefit from relationship discovery.
+
+## Legacy Compatibility
+
+v0.6.x projects do not need migration. If `.pmem/manifest.yml` has no `schema` block, pmem falls back to the legacy software defaults and does not rewrite the manifest.