memorix 1.7.2 → 1.8.0

package/docs/1.2.4-PERSISTENT-MEMORY-DELIVERY.md

@@ -26,7 +26,7 @@ database, a transcript archive, or another default MCP tool.
 | Start a new, unrelated task | A bounded task Workset. No automatic prior-session text. |
 | Continue or take over prior work | The same Workset plus a bounded prior-work projection. |
 | MCP is available | Use the existing `memorix_project_context` tool with the real task. |
-| MCP is unavailable | Make one CLI call: `memorix resume "<task>" --brief-json` for continuation or `memorix context "<task>" --brief-json` for a new task. `--json` remains the detailed diagnostic form. |
+| MCP is unavailable | Make one CLI fallback call: `memorix resume "<task>" --fallback --brief-json` for continuation or `memorix context "<task>" --fallback --brief-json` for a new task. The fallback budget rejects repeated probes; `--json` remains the detailed diagnostic form. |
 | Claude Code hook is installed | An explicit continuation prompt receives the same bounded prior-work Workset through the official `UserPromptSubmit` context channel, unless `memory.inject = "silent"`. |
 | Read-only assessment | The brief may be read, but Memorix does not persist a new observation unless the user explicitly asks to save one. |
 | Need a deliberate session handoff | Keep using the explicit session/handoff surfaces; this projection does not replace them. |

package/docs/1.8.0-RELEASE-CANDIDATE-SPEC.md

@@ -0,0 +1,75 @@
+# Memorix 1.8.0 Release Candidate Specification
+
+Status: accepted implementation contract for Memorix 1.8.0.
+
+## Acceptance Checklist
+
+- [x] Keep one canonical release document and update the documentation map.
+- [x] Expose MCP protocol diagnostics without changing existing defaults.
+- [x] Preserve legacy stateful Streamable HTTP and stdio clients.
+- [x] Return bounded evidence cards from persisted project graph state.
+- [x] Provide a deterministic memory feedback reducer, SQLite event persistence,
+  auditable MCP/CLI adapters, and a unified retrieval/retention projection.
+- [x] Keep the default MCP profile compact and preserve the full profile contract.
+- [x] Provide dashboard evidence/provenance views without replacing the existing UI.
+- [x] Preserve explicit project/session handles across restart and client shapes.
+- [x] Verify CLI fallback, Windows child-process hiding, HTTP old/new smoke, build, and package contents.
+
+## Compatibility Contract
+
+The MCP 2026-07-28 protocol is the external stateless request contract.
+Memorix also accepts the legacy stateful Streamable HTTP contract with
+`Mcp-Session-Id`. Stateless requests carry an explicit `Mcp-Project-Handle`
+that is persisted in SQLite and survives service restart; the handle, rather
+than a transport session id, owns project routing. The pinned SDK's accepted
+wire version is normalized internally at the HTTP boundary, while responses
+continue to advertise current protocol metadata. Unsupported Tasks/polling is
+reported by capability diagnostics rather than simulated.
+
+Antigravity is the primary setup target. Gemini CLI remains a compatibility and
+migration path. GitHub issue #204 (WorkBuddy) is intentionally not merged or
+used as release evidence.
+
+## Evidence Architecture
+
+Consumers are both humans and agents. Markdown remains the human source of
+truth; SQLite graph records and bounded evidence cards are the agent index.
+Retrieval is hybrid/graph-assisted, but never a full graph dump. Supported
+relations are persisted through the existing graph store and may reference
+files, symbols, calls/imports, tests, scripts, commits, dependencies,
+failures, memory, and source. Non-TypeScript/JavaScript projects use the
+existing graceful fallback and report coverage honestly.
+
+## Memory Feedback Contract
+
+The lifecycle is capture -> qualify -> recall -> use -> feedback ->
+strengthen/weaken -> archive/forget. Feedback is auditable and deterministic.
+Supported signals are user correction, verification success/failure, code
+change, use/non-use, and conflict with current code. Positive feedback raises
+retrieval weight; negative or conflicting feedback lowers it and can make a
+record eligible for archive. Recovery is explicit and auditable.
+
+The 1.8.0 release ships the reducer, durable state/event tables, MCP and CLI
+surfaces, revoke events, and tests for these rules. A legacy record without a
+feedback event keeps its historical retrieval/retention score; the first
+explicit feedback event opts it into the feedback projection.
+
+## Evidence and Operator Surface
+
+Every observation write creates or refreshes a project-scoped evidence card with
+source kind, source reference, locator, captured hash, files, and freshness.
+Changed paths can mark cards stale without deleting their source memory. The
+`memorix_evidence` MCP tool, `memorix evidence` CLI command, and Dashboard
+Evidence page expose the same persisted cards and audit events.
+
+When MCP discovery is unavailable, `memorix context --fallback` uses a durable
+one-shot budget keyed by project and task. Repeated help/search/status probing
+is rejected for that task window.
+
+## Explicit Non-goals for 1.8.0
+
+- No fake MCP Tasks/polling implementation when the pinned SDK does not expose
+  a durable server-side task contract.
+- No remote memory hub, proxying of model traffic, or upstream installer.
+- No claim that the bundled Lite parser is a semantic CodeGraph.
+- No marketing landing page in place of an operator evidence console.

package/docs/API_REFERENCE.md

@@ -103,7 +103,7 @@ memorix context "continue auth bug"
 memorix resume "continue auth bug"
 memorix checkpoint list
 memorix checkpoint context --task "continue auth bug"
-memorix context --task "prepare 1.1.7 release"
+  memorix context --task "prepare 1.8.0 release"
 memorix explain
 memorix codegraph context-pack --task "continue auth bug"
 ```
@@ -123,6 +123,42 @@ SessionStart hooks keep the default minimal hint lightweight. When memory behavi
 
 The intended loop for agents is: get the project brief when it helps, inspect the suggested current files, use stale or unbound memory only as a lead, store durable outcomes after the work changes the project, and resolve obsolete memories.
 
+### Evidence Cards and Feedback
+
+Evidence Cards are the persisted provenance index for project memory. They do
+not replace the observation, source file, commit, or test they point to. Every
+observation write creates or refreshes a card with `sourceRef`, `locator`,
+`capturedHash`, project ownership, files, verification state, and freshness.
+
+CLI:
+
+```bash
+memorix evidence list --json
+memorix evidence get --id 42 --json
+memorix evidence stale --paths src/server.ts,tests/server.test.ts --json
+memorix feedback record --id 42 --signal verification-success --source test:run-1 --json
+memorix feedback audit --id 42 --json
+```
+
+MCP uses `memorix_evidence` for `list`, `get`, `sync`, `stale`, and `events`,
+and `memorix_feedback` for `record`, `show`, and `audit`. Feedback events are
+append-only; `revoke` points at the event being undone, so the audit history is
+preserved. A memory without explicit feedback keeps the legacy relevance and
+retention score. The first feedback event opts that candidate into the
+feedback projection.
+
+### HTTP MCP Compatibility
+
+Legacy clients use stateful Streamable HTTP with `Mcp-Session-Id`. Current
+stateless clients send `Mcp-Protocol-Version: 2026-07-28` and receive a durable
+`Mcp-Project-Handle`. Send that handle on subsequent requests; it is persisted
+in SQLite and remains valid across service restart until its TTL expires.
+`/protocol-diagnostics` reports the current/legacy contract and explicitly
+reports Tasks and polling as unsupported when the pinned SDK cannot provide a
+durable implementation. When MCP discovery is unavailable, use one bounded
+`memorix context --fallback` request for a task; repeated fallback probing is
+rejected by the local budget.
+
 The built-in Lite provider indexes common code files with lightweight file, symbol, and import facts. It is a structural fallback, not a language-server-quality graph. When a project already has a healthy local CodeGraph index, `[codegraph].external_context = "auto"` may add a validated, bounded semantic outline to the Workset. Memorix never initializes, syncs, or exports that external index, and never stores its raw source output. `memorix codegraph status --json`, `memorix doctor --json`, and Project Context JSON identify the actual provider quality.
 
 ### Knowledge Workspace and Workflows

package/docs/ARCHITECTURE.md

@@ -391,7 +391,7 @@ Resolution order:
 2. project `.env`
 3. user `~/.memorix/.env`
 
-The primary lanes are `[agent]` for memcode, `[memory.llm]` for formation/rerank/summaries, and `[embedding]` for semantic search. The dashboard and `memorix status` expose config provenance so the active value source is visible.
+The primary lanes are `[agent]` for memcode, `[memory.llm]` for formation/summaries/LLM-rerank fallback, `[rerank]` for optional HTTP rerank against a compatible `/rerank` API, and `[embedding]` for semantic search. The dashboard and `memorix status` expose config provenance so the active value source is visible.
 
 ---
 

package/docs/CONFIGURATION.md

@@ -47,6 +47,12 @@ model = "text-embedding-v4"
 base_url = "https://dashscope.aliyuncs.com/compatible-mode/v1"
 api_key = "..."
 
+# Optional HTTP rerank. Off unless provider = "http".
+# [rerank]
+# provider = "http"
+# model = "rerank-model"
+# base_url = "https://api.example.com/v1"
+
 [memory]
 inject = "minimal"
 formation = "active"
@@ -127,7 +133,7 @@ Used by Memorix background memory intelligence:
 - memory formation
 - summarization
 - deduplication
-- optional reranking
+- optional LLM reranking fallback
 - cleanup assistance
 
 Common keys:
@@ -139,6 +145,9 @@ Common keys:
 
 For OpenAI-compatible providers such as DashScope, DeepSeek-compatible gateways,
 or internal model gateways, use `provider = "openai"` and set `base_url`.
+Memorix uses an explicit provider path as-is: for example,
+`https://ark.cn-beijing.volces.com/api/v3` remains `/api/v3`. Only a bare host
+without a versioned path receives the convenience suffix `/v1`.
 
 ### `[embedding]`
 
@@ -186,6 +195,53 @@ OpenAI-only default. The equivalent env-var form is
 `MEMORIX_EMBEDDING=api`, `MEMORIX_EMBEDDING_BASE_URL=https://openrouter.ai/api/v1`,
 `MEMORIX_EMBEDDING_MODEL=qwen/qwen3-embedding-8b`.
 
+### `[rerank]`
+
+Optional HTTP rerank for thorough search. Off by default.
+When `provider = "http"`, Memorix POSTs a Cohere-compatible
+`{model, query, documents}` body to `{base_url}/rerank` (or an explicit
+`/rerank` path). Any compatible endpoint works — hosted APIs, TEI, vLLM,
+or a local gateway. Memorix sends the configured `model` string unchanged.
+
+Common keys:
+
+- `provider` — `off` (default) or `http`
+- `model` — user-configured model id (no built-in default). Example:
+  `rerank-model`
+- `base_url` — API root that serves `/rerank`, for example
+  `https://api.example.com/v1` or `http://127.0.0.1:8080`. When unset and
+  `provider = "http"`, Memorix inherits `[memory.llm].base_url`.
+- `api_key` — bearer token. When unset, inherits the memory LLM key
+  (`MEMORIX_LLM_API_KEY`, `MEMORIX_API_KEY`, or `[memory.llm].api_key`)
+  only if the effective `base_url` is the same trusted endpoint as
+  `[memory.llm].base_url`. A different rerank URL requires
+  `MEMORIX_RERANK_API_KEY` or `rerank.api_key`.
+
+Environment overrides (highest priority after CLI flags):
+
+```bash
+MEMORIX_RERANK_PROVIDER=http
+MEMORIX_RERANK_MODEL=rerank-model
+MEMORIX_RERANK_BASE_URL=https://api.example.com/v1
+# MEMORIX_RERANK_API_KEY=   # required when base_url is not the memory LLM endpoint
+```
+
+If the memory LLM lane already points at the same gateway, you can enable
+rerank with:
+
+```toml
+[rerank]
+provider = "http"
+model = "rerank-model"
+```
+
+or set `MEMORIX_RERANK_PROVIDER=http` and `MEMORIX_RERANK_MODEL=...`.
+Timeout is `MEMORIX_RERANK_TIMEOUT_MS` (default 30000). Thorough search
+with at least 3 candidates uses HTTP rerank when configured. On HTTP miss
+or timeout, Memorix keeps the original order and does not fall through to
+LLM rerank. LLM rerank remains available when HTTP rerank is off, on the
+existing thorough/heavy/ambiguous gate.
+
 Image analysis (visual description for ingested images) runs on the LLM lane
 through an OpenAI-compatible vision endpoint, so it can also use OpenRouter
 with a multimodal model — for example `qwen/qwen3-vl-8b-instruct` via

package/docs/DASHBOARD.md

@@ -2,6 +2,11 @@
 
 The Memorix Dashboard is the local control surface for inspecting project memory and running explicit maintenance. It is available through the shared background service or as a standalone local process.
 
+The 1.8.0 Evidence page is the operator view of the same source-backed cards
+used by MCP and CLI. It shows each card's source reference, locator, captured
+hash, freshness, verification state, and audit-event count. It is project
+scoped and does not replace the underlying observation or code source.
+
 ## Start
 
 ```bash
@@ -52,3 +57,8 @@ POST /api/maintenance/retention/execute
 ```
 
 Use the CLI for scripts and remote terminals. Use MCP for agent-driven maintenance. The Dashboard is the reviewable local operator surface; all three call the same maintenance logic.
+
+The control-plane API also exposes `GET /api/evidence?project=<id>` for the
+bounded Evidence Console payload. The page is responsive at desktop, tablet,
+and narrow mobile widths; long paths wrap instead of widening the viewport,
+and maintenance dialogs remain viewport-centered.

package/docs/DEVELOPMENT.md

@@ -13,8 +13,8 @@ Memorix is a TypeScript project built around:
 
 ## Current Development Baseline
 
-The current release work targets the **1.7 line** while package metadata may
-still show the last published version until the release commit is cut.
+The current release work targets the **1.8.0 release line** and package metadata
+is kept at `1.8.0` for this release commit.
 
 Contributors should assume the following areas are part of the current release line:
 
@@ -29,6 +29,7 @@ Contributors should assume the following areas are part of the current release l
 - the existing layered retrieval, retention, attribution, and compact output model
 - source-backed long-term memory with candidate, qualification, approval, archival, supersession, and explicit portable user scope
 - preview-first project maintenance shared across Dashboard, CLI, and MCP
+- persisted Evidence Cards, feedback audit events, and stateless MCP project handles
 - the reproducible large-store gate for SDK, HTTP MCP, hooks, persistence, and
   cold-start retrieval budgets
 

package/docs/INTEGRATIONS.md

@@ -61,7 +61,7 @@ Generated guidance also has scope:
 | Pi | `memorix setup --agent pi --global` | Pi package | User Pi package with extension and official skills, registered through `pi install <path> --approve` | Run without `--global` only when you want a project-local Pi package. Pi currently does not need a separate Memorix MCP config lane. |
 | Windsurf | `memorix setup --agent windsurf --global` | Windsurf MCP/rules/hooks config | stdio MCP config, `.windsurf/rules/memorix.md`, hook config | Uses Windsurf's current config surfaces. |
 | Kiro | `memorix setup --agent kiro --global` | Kiro MCP/steering/hooks config | MCP config, `.kiro/steering/memorix.md`, `.kiro/hooks/*.kiro.hook` | Uses Kiro steering and hook files. |
-| Antigravity | `memorix setup --agent antigravity --global` | Antigravity plugin | Plugin under `~/.gemini/config/plugins/memorix` or `.agents/plugins/memorix`, with `plugin.json`, `mcp_config.json`, `hooks.json`, rules, and skills | Uses Antigravity's official plugin layout. Dedicated MCP configs live at `~/.gemini/config/mcp_config.json` or `.agents/mcp_config.json`; legacy Gemini settings are read only for compatibility. |
+| Antigravity | `memorix setup --agent antigravity --global` | **Primary supported target** | Plugin under `~/.gemini/config/plugins/memorix` or `.agents/plugins/memorix`, with `plugin.json`, `mcp_config.json`, `hooks.json`, rules, and skills | Uses Antigravity's official plugin layout. Dedicated MCP configs live at `~/.gemini/config/mcp_config.json` or `.agents/mcp_config.json`; legacy Gemini settings are read only for compatibility. |
 | Trae | `memorix setup --agent trae --global` | Trae MCP/rules config | MCP config and `.trae/rules/project_rules.md` | Current support is MCP plus project rules. |
 | DeepSeek Harness | `memorix setup --agent dsh --global` | DeepSeek Harness MCP patch | Memorix `@deepseek-ai/dsh-mcp-client` row in `$DSH_HOME/cordis.patch.yml` (default `~/.dsh/cordis.patch.yml`), guidance in the harness `AGENTS.md`, and official skills under `$DSH_HOME/skills` | Tools appear as `mcp__memorix__*`. No hooks surface. Project-local setup writes project `AGENTS.md` guidance and `.dsh/skills`; the MCP row always lives in the harness-level patch file. |
 | memcode | `memorix` or `memcode` | Bundled terminal agent | Built-in Memorix memory access | memcode uses the same project memory pool; it is not a separate memory silo. |

package/docs/PERFORMANCE.md

@@ -75,7 +75,9 @@ On the release development machine used for this check, the healthy HTTP service
 | `MEMORIX_FORMATION_TIMEOUT_MS` | `12000` (12 s) | Raise when LLM-backed formation should outlive slow proxy/provider hops |
 | `MEMORIX_LLM_API_KEY` / `OPENAI_API_KEY` | unset | Enable LLM-backed enrichment, extraction, rerank, or skill generation |
 | `MEMORIX_LLM_TIMEOUT_MS` | `30000` (30 s) | Bound a single LLM-backed extraction/resolve call |
-| `MEMORIX_RERANK_TIMEOUT_MS` | provider default | Bound slow LLM rerank calls |
+| `MEMORIX_RERANK_TIMEOUT_MS` | 30000 | Bound HTTP and LLM rerank calls |
+| `MEMORIX_RERANK_PROVIDER` | `off` | Set `http` to enable optional HTTP rerank |
+| `MEMORIX_RERANK_BASE_URL` | `[memory.llm].base_url` | Compatible `/rerank` API root (path `/rerank` is appended) |
 | `memorix memory search --quality fast` | n/a | Force a fully local retrieval path for a latency-sensitive call |
 | `npm run benchmark:retrieval -- --records 1000 --runs 100` | n/a | Reproduce hot in-process lexical retrieval latency; not an end-to-end claim |
 | `npm run gate:large-store -- --records 40000` | n/a | Exercise SDK, HTTP MCP, hook persistence, cache integrity, and reopen behavior against a large isolated store |

package/docs/README.md

@@ -37,6 +37,7 @@ The public docs are organized by user intent:
 | Reviewable knowledge pages and workflow inheritance | [API_REFERENCE.md § Knowledge Workspace and Workflows](API_REFERENCE.md#knowledge-workspace-and-workflows), [1.2 Knowledge Workspace](1.2.0-KNOWLEDGE-WORKSPACE.md), and [1.2 Workflow Inheritance](1.2.0-WORKFLOW-INHERITANCE.md) |
 | Git-derived engineering memory | [GIT_MEMORY.md](GIT_MEMORY.md) |
 | Memory formation and quality pipeline | [MEMORY_FORMATION_PIPELINE.md](MEMORY_FORMATION_PIPELINE.md) |
+| Evidence cards and memory feedback | [1.8.0 Release Specification](1.8.0-RELEASE-CANDIDATE-SPEC.md) and [API_REFERENCE.md](API_REFERENCE.md) |
 | Performance and resource profile | [PERFORMANCE.md](PERFORMANCE.md) |
 | Orchestrated subagent tasks, messages, locks, handoffs | [API_REFERENCE.md § Orchestration Coordination](API_REFERENCE.md#9-orchestration-coordination-tools) |
 | Subagent orchestration | [API_REFERENCE.md](API_REFERENCE.md) and `memorix orchestrate --help` |
@@ -92,6 +93,7 @@ The public docs are organized by user intent:
 | 1.2 honest Lite and optional semantic CodeGraph provider contract | [1.2 Provider Quality](1.2.0-PROVIDER-QUALITY.md) |
 | 1.3 long-term memory model, source boundary, and lifecycle | [1.3 Memory Architecture](1.3-MEMORY-ARCHITECTURE.md) |
 | Active context-control work | [1.2.2 Memory Control Plane](1.2.2-MEMORY-CONTROL-PLANE.md) |
+| 1.8 context-control acceptance contract | [1.8.0 Release Specification](1.8.0-RELEASE-CANDIDATE-SPEC.md) |
 | Historical cloud sync and multi-agent research | [CLOUD_SYNC_AND_MULTI_AGENT_RESEARCH.md](CLOUD_SYNC_AND_MULTI_AGENT_RESEARCH.md) |
 | Known issues and old roadmap notes | [KNOWN_ISSUES_AND_ROADMAP.md](KNOWN_ISSUES_AND_ROADMAP.md) |
 
@@ -119,7 +121,9 @@ Historical/deep-reference documents may describe older designs. If they conflict
 
 ## Current Product Line
 
-The released product is on the **1.7 line**. The current baseline has:
+The current product line is **1.8.0**. The authoritative acceptance contract is
+[1.8.0 Release Specification](1.8.0-RELEASE-CANDIDATE-SPEC.md).
+The current baseline has:
 
 - `memorix setup --agent <agent> --global` is the default agent integration command
 - `memorix serve` is the manual stdio MCP server for external agents
@@ -132,3 +136,9 @@ The released product is on the **1.7 line**. The current baseline has:
 - `memorix` / `memcode` open memcode, the bundled terminal agent that uses the same Memorix memory pool
 - `~/.memorix/config.toml` and project `memorix.toml` are the user-facing configuration model
 - legacy `memorix.yml`, `.env`, and `config.json` files are still read for compatibility, but new setups should use TOML
+
+For MCP compatibility, 1.8 preserves legacy stateful Streamable HTTP and
+supports 2026-07-28 stateless requests with durable project handles. Tasks and
+polling are reported as unsupported when the pinned SDK cannot provide a real
+durable implementation. Antigravity is the primary setup target; Gemini CLI is
+the compatibility path. See the release specification for the exact contract.

package/docs/SETUP.md

@@ -498,7 +498,8 @@ What each lane does:
 | Lane | Used by |
 | --- | --- |
 | `[agent]` | model used by memcode while coding |
-| `[memory.llm]` | memory formation, summaries, deduplication, optional rerank |
+| `[memory.llm]` | memory formation, summaries, deduplication, optional LLM rerank fallback |
+| `[rerank]` | optional HTTP rerank against a compatible `/rerank` API |
 | `[embedding]` | semantic/vector search |
 | `[memory]` | memory injection and formation behavior |
 | `[git]` | Git Memory hook and ingestion behavior |

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "memorix",
   "mcpName": "io.github.AVIDS2/memorix",
-  "version": "1.7.2",
+  "version": "1.8.0",
   "description": "Local-first shared memory layer for AI coding agents across MCP clients, Git history, reasoning context, and project sessions.",
   "type": "module",
   "sideEffects": false,

package/plugins/antigravity/memorix/rules/memorix.md

@@ -3,7 +3,7 @@
 Use Memorix when prior workspace context, decisions, fixes, or handoff state would materially help the current task and Memorix tools are available.
 
 - Call `memorix_project_context` with the user's actual task for continuation, fresh handoff, or unfamiliar coding work so Memorix can choose a task-lensed brief.
-- If the MCP tool is not visible yet but the client supports tool discovery or dynamic loading, search/select `memorix_project_context` first. Run `memorix context --task "<task>"` only after MCP is unavailable, disabled, or not discoverable, and pass the user's real task text.
+- If the MCP tool is not visible yet but the client supports tool discovery or dynamic loading, search/select `memorix_project_context` first. Run `memorix context --task "<task>" --fallback --brief-json` only after MCP is unavailable, disabled, or not discoverable, and pass the user's real task text.
 - Use `memorix_context_pack` when you need structured refs and freshness for code-bound memories.
 - Use `memorix_graph_context` for explicit memory graph questions.
 - Use `memorix_search` for focused lookup.

package/plugins/claude/memorix/.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "$schema": "https://anthropic.com/claude-code/plugin.schema.json",
   "name": "memorix",
-  "version": "1.7.2",
+  "version": "1.8.0",
   "description": "Shared workspace memory for Claude Code and other AI coding agents.",
   "author": {
     "name": "AVIDS2",

package/plugins/claude/memorix/SKILL.md

@@ -32,7 +32,7 @@ is not a failure by itself. Claude Code supports MCP tool search, and the
 Memorix Claude plugin asks Claude to eagerly load its small core tool set. In
 headless/print-mode, if Memorix tools are not in the first visible tool list,
 search/select `memorix_project_context` before using shell fallback. Run
-`memorix context --task "<task>"` only after MCP is unavailable, disabled, or
+`memorix context --task "<task>" --fallback --brief-json` only after MCP is unavailable, disabled, or
 not discoverable, and pass the user's real task text.
 
 Do not store secrets, credentials, raw private transcripts, or trivial one-off actions. Treat memory as context, not as a substitute for reading code and verifying behavior.

package/plugins/claude/memorix/skills/memorix-memory/SKILL.md

@@ -11,7 +11,7 @@ Use Memorix as the shared memory layer for the active workspace when Memorix too
 
 | Situation | Prefer | CLI fallback |
 |---|---|---|
-| Broad continuation, fresh handoff, or "what do we know?" | `memorix_project_context` with the user's task | `memorix context --task "<topic>"` |
+| Broad continuation, fresh handoff, or "what do we know?" | `memorix_project_context` with the user's task | `memorix context --task "<topic>" --fallback --brief-json` |
 | Need structured refs/freshness for code-bound memories | `memorix_context_pack` | `memorix codegraph context-pack --task "<topic>"` |
 | Specific past decision, bug, file, or change | `memorix_search` | `memorix memory search --query "<topic>"` |
 | Need the full source for a search hit | `memorix_detail` | `memorix memory detail --id <id>` |
@@ -28,7 +28,7 @@ Use Memorix as the shared memory layer for the active workspace when Memorix too
 - Treat memory as background context. Still read the current code and verify behavior.
 - Skip memory lookup for greetings, tiny one-off edits, or questions fully answered by the current file.
 - If a fresh project has no memories, proceed normally and do not repeat the same empty search in the same turn.
-- If Claude Code print-mode shows Memorix MCP as pending and the tools are not visible yet, first use Claude Code's MCP tool search/dynamic loading to find `memorix_project_context`. Run `memorix context --task "<task>"` from the shell only after MCP is unavailable, disabled, or not discoverable, and pass the user's real task text. Do not skip memory, wait indefinitely on `pending`, or hand-write tool-call syntax.
+- If Claude Code print-mode shows Memorix MCP as pending and the tools are not visible yet, first use Claude Code's MCP tool search/dynamic loading to find `memorix_project_context`. Run `memorix context --task "<task>" --fallback --brief-json` from the shell only after MCP is unavailable, disabled, or not discoverable, and pass the user's real task text. Do not skip memory, wait indefinitely on `pending`, or hand-write tool-call syntax.
 
 ## Store Rules
 

package/plugins/codebuddy/memorix/.codebuddy-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "memorix",
-  "version": "1.7.2",
+  "version": "1.8.0",
   "description": "Local-first project memory for CodeBuddy Code and other coding agents.",
   "author": {
     "name": "AVIDS2",

package/plugins/codebuddy/memorix/skills/memorix-memory/SKILL.md

@@ -11,7 +11,7 @@ Use Memorix as the shared memory layer for the active workspace when Memorix too
 
 | Situation | Prefer | CLI fallback |
 |---|---|---|
-| Broad continuation, fresh handoff, or "what do we know?" | `memorix_project_context` with the user's task | `memorix context --task "<topic>"` |
+| Broad continuation, fresh handoff, or "what do we know?" | `memorix_project_context` with the user's task | `memorix context --task "<topic>" --fallback --brief-json` |
 | Need structured refs/freshness for code-bound memories | `memorix_context_pack` | `memorix codegraph context-pack --task "<topic>"` |
 | Specific past decision, bug, file, or change | `memorix_search` | `memorix memory search --query "<topic>"` |
 | Need the full source for a search hit | `memorix_detail` | `memorix memory detail --id <id>` |
@@ -28,7 +28,7 @@ Use Memorix as the shared memory layer for the active workspace when Memorix too
 - Treat memory as background context. Still read the current code and verify behavior.
 - Skip memory lookup for greetings, tiny one-off edits, or questions fully answered by the current file.
 - If a fresh project has no memories, proceed normally and do not repeat the same empty search in the same turn.
-- If Claude Code print-mode shows Memorix MCP as pending and the tools are not visible yet, first use Claude Code's MCP tool search/dynamic loading to find `memorix_project_context`. Run `memorix context --task "<task>"` from the shell only after MCP is unavailable, disabled, or not discoverable, and pass the user's real task text. Do not skip memory, wait indefinitely on `pending`, or hand-write tool-call syntax.
+- If Claude Code print-mode shows Memorix MCP as pending and the tools are not visible yet, first use Claude Code's MCP tool search/dynamic loading to find `memorix_project_context`. Run `memorix context --task "<task>" --fallback --brief-json` from the shell only after MCP is unavailable, disabled, or not discoverable, and pass the user's real task text. Do not skip memory, wait indefinitely on `pending`, or hand-write tool-call syntax.
 
 ## Store Rules
 

package/plugins/codex/memorix/.codex-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "memorix",
-  "version": "1.7.2",
+  "version": "1.8.0",
   "description": "Shared workspace memory for Codex and other AI coding agents.",
   "author": {
     "name": "AVIDS2",

package/plugins/codex/memorix/skills/memorix-memory/SKILL.md

@@ -11,7 +11,7 @@ Use Memorix as the shared memory layer for the active workspace when Memorix too
 
 | Situation | Prefer | CLI fallback |
 |---|---|---|
-| Broad continuation, fresh handoff, or "what do we know?" | `memorix_project_context` with the user's task | `memorix context --task "<topic>"` |
+| Broad continuation, fresh handoff, or "what do we know?" | `memorix_project_context` with the user's task | `memorix context --task "<topic>" --fallback --brief-json` |
 | Need structured refs/freshness for code-bound memories | `memorix_context_pack` | `memorix codegraph context-pack --task "<topic>"` |
 | Explicit memory graph question | `memorix_graph_context` | `memorix memory graph-context --query "<topic>"` |
 | Specific past decision, bug, file, or change | `memorix_search` | `memorix memory search --query "<topic>"` |
@@ -29,7 +29,7 @@ Use Memorix as the shared memory layer for the active workspace when Memorix too
 - Treat memory as background context. Still read the current code and verify behavior.
 - Skip memory lookup for greetings, tiny one-off edits, or questions fully answered by the current file.
 - If a fresh project has no memories, proceed normally and do not repeat the same empty search in the same turn.
-- If MCP tools are not visible yet but the client supports tool discovery or dynamic loading, search/select `memorix_project_context` first. Run `memorix context --task "<task>"` only after MCP is unavailable, disabled, or not discoverable, and pass the user's real task text. Do not skip memory, wait indefinitely on startup, or hand-write tool-call syntax.
+- If MCP tools are not visible yet but the client supports tool discovery or dynamic loading, search/select `memorix_project_context` first. Run `memorix context --task "<task>" --fallback --brief-json` only after MCP is unavailable, disabled, or not discoverable, and pass the user's real task text. Do not skip memory, wait indefinitely on startup, or hand-write tool-call syntax.
 
 ## Store Rules
 

package/plugins/copilot/memorix/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "memorix",
-  "version": "1.7.2",
+  "version": "1.8.0",
   "description": "Shared local workspace memory for GitHub Copilot CLI and other AI coding agents.",
   "author": {
     "name": "AVIDS2",

package/plugins/copilot/memorix/skills/memorix-memory/SKILL.md

@@ -11,7 +11,7 @@ Use Memorix as the shared memory layer for the active workspace when Memorix too
 
 | Situation | Prefer | CLI fallback |
 |---|---|---|
-| Broad continuation, fresh handoff, or "what do we know?" | `memorix_project_context` with the user's task | `memorix context --task "<topic>"` |
+| Broad continuation, fresh handoff, or "what do we know?" | `memorix_project_context` with the user's task | `memorix context --task "<topic>" --fallback --brief-json` |
 | Need structured refs/freshness for code-bound memories | `memorix_context_pack` | `memorix codegraph context-pack --task "<topic>"` |
 | Explicit memory graph question | `memorix_graph_context` | `memorix memory graph-context --query "<topic>"` |
 | Specific past decision, bug, file, or change | `memorix_search` | `memorix memory search --query "<topic>"` |
@@ -29,7 +29,7 @@ Use Memorix as the shared memory layer for the active workspace when Memorix too
 - Treat memory as background context. Still read the current code and verify behavior.
 - Skip memory lookup for greetings, tiny one-off edits, or questions fully answered by the current file.
 - If a fresh project has no memories, proceed normally and do not repeat the same empty search in the same turn.
-- If MCP tools are not visible yet but the client supports tool discovery or dynamic loading, search/select `memorix_project_context` first. Run `memorix context --task "<task>"` only after MCP is unavailable, disabled, or not discoverable, and pass the user's real task text. Do not skip memory, wait indefinitely on startup, or hand-write tool-call syntax.
+- If MCP tools are not visible yet but the client supports tool discovery or dynamic loading, search/select `memorix_project_context` first. Run `memorix context --task "<task>" --fallback --brief-json` only after MCP is unavailable, disabled, or not discoverable, and pass the user's real task text. Do not skip memory, wait indefinitely on startup, or hand-write tool-call syntax.
 
 ## Store Rules
 

package/plugins/cursor/memorix/skills/memorix-memory/SKILL.md

@@ -11,7 +11,7 @@ Use Memorix as the shared memory layer for the active workspace when Memorix too
 
 | Situation | Prefer | CLI fallback |
 |---|---|---|
-| Broad continuation, fresh handoff, or "what do we know?" | `memorix_project_context` with the user's task | `memorix context --task "<topic>"` |
+| Broad continuation, fresh handoff, or "what do we know?" | `memorix_project_context` with the user's task | `memorix context --task "<topic>" --fallback --brief-json` |
 | Need structured refs/freshness for code-bound memories | `memorix_context_pack` | `memorix codegraph context-pack --task "<topic>"` |
 | Explicit memory graph question | `memorix_graph_context` | `memorix memory graph-context --query "<topic>"` |
 | Specific past decision, bug, file, or change | `memorix_search` | `memorix memory search --query "<topic>"` |
@@ -29,7 +29,7 @@ Use Memorix as the shared memory layer for the active workspace when Memorix too
 - Treat memory as background context. Still read the current code and verify behavior.
 - Skip memory lookup for greetings, tiny one-off edits, or questions fully answered by the current file.
 - If a fresh project has no memories, proceed normally and do not repeat the same empty search in the same turn.
-- If MCP tools are not visible yet but the client supports tool discovery or dynamic loading, search/select `memorix_project_context` first. Run `memorix context --task "<task>"` only after MCP is unavailable, disabled, or not discoverable, and pass the user's real task text. Do not skip memory, wait indefinitely on startup, or hand-write tool-call syntax.
+- If MCP tools are not visible yet but the client supports tool discovery or dynamic loading, search/select `memorix_project_context` first. Run `memorix context --task "<task>" --fallback --brief-json` only after MCP is unavailable, disabled, or not discoverable, and pass the user's real task text. Do not skip memory, wait indefinitely on startup, or hand-write tool-call syntax.
 
 ## Store Rules
 

package/plugins/gemini/memorix/GEMINI.md

@@ -5,7 +5,7 @@ This extension adds Memorix memory guidance for the active Gemini CLI workspace.
 Use Memorix when prior workspace context, decisions, fixes, or handoff state would materially help the current task and Memorix tools are available.
 
 - Call `memorix_project_context` with the user's actual task for continuation, fresh handoff, or unfamiliar coding work so Memorix can choose a task-lensed brief.
-- If the MCP tool is not visible yet but the client supports tool discovery or dynamic loading, search/select `memorix_project_context` first. Run `memorix context --task "<task>"` only after MCP is unavailable, disabled, or not discoverable, and pass the user's real task text.
+- If the MCP tool is not visible yet but the client supports tool discovery or dynamic loading, search/select `memorix_project_context` first. Run `memorix context --task "<task>" --fallback --brief-json` only after MCP is unavailable, disabled, or not discoverable, and pass the user's real task text.
 - Use `memorix_context_pack` when you need structured refs and freshness for code-bound memories.
 - Use `memorix_graph_context` for explicit memory graph questions.
 - Use `memorix_search` for focused lookup.

package/plugins/gemini/memorix/gemini-extension.json

@@ -1,6 +1,6 @@
 {
   "name": "memorix",
-  "version": "1.7.2",
+  "version": "1.8.0",
   "mcpServers": {
     "memorix": {
       "command": "memorix",

package/plugins/omp/memorix/package.json

@@ -1,6 +1,6 @@
 {
   "name": "memorix-omp-package",
-  "version": "1.7.2",
+  "version": "1.8.0",
   "description": "Memorix shared workspace memory package for Oh-my-Pi.",
   "license": "Apache-2.0",
   "keywords": [

package/plugins/openclaw/memorix/.codex-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "memorix",
-  "version": "1.7.2",
+  "version": "1.8.0",
   "description": "Shared workspace memory for OpenClaw and other AI coding agents.",
   "author": {
     "name": "AVIDS2",

package/plugins/pi/memorix/package.json

@@ -1,6 +1,6 @@
 {
   "name": "memorix-pi-package",
-  "version": "1.7.2",
+  "version": "1.8.0",
   "description": "Memorix shared workspace memory package for Pi coding agent.",
   "license": "Apache-2.0",
   "keywords": [

package/plugins/pi/memorix/skills/memorix-memory/SKILL.md

@@ -11,7 +11,7 @@ Use Memorix as the shared memory layer for the active workspace when Memorix too
 
 | Situation | Prefer | CLI fallback |
 |---|---|---|
-| Broad continuation, fresh handoff, or "what do we know?" | `memorix_project_context` with the user's task | `memorix context --task "<topic>"` |
+| Broad continuation, fresh handoff, or "what do we know?" | `memorix_project_context` with the user's task | `memorix context --task "<topic>" --fallback --brief-json` |
 | Need structured refs/freshness for code-bound memories | `memorix_context_pack` | `memorix codegraph context-pack --task "<topic>"` |
 | Explicit memory graph question | `memorix_graph_context` | `memorix memory graph-context --query "<topic>"` |
 | Specific past decision, bug, file, or change | `memorix_search` | `memorix memory search --query "<topic>"` |
@@ -29,7 +29,7 @@ Use Memorix as the shared memory layer for the active workspace when Memorix too
 - Treat memory as background context. Still read the current code and verify behavior.
 - Skip memory lookup for greetings, tiny one-off edits, or questions fully answered by the current file.
 - If a fresh project has no memories, proceed normally and do not repeat the same empty search in the same turn.
-- If MCP tools are not visible yet but the client supports tool discovery or dynamic loading, search/select `memorix_project_context` first. Run `memorix context --task "<task>"` only after MCP is unavailable, disabled, or not discoverable, and pass the user's real task text. Do not skip memory, wait indefinitely on startup, or hand-write tool-call syntax.
+- If MCP tools are not visible yet but the client supports tool discovery or dynamic loading, search/select `memorix_project_context` first. Run `memorix context --task "<task>" --fallback --brief-json` only after MCP is unavailable, disabled, or not discoverable, and pass the user's real task text. Do not skip memory, wait indefinitely on startup, or hand-write tool-call syntax.
 
 ## Store Rules
 

package/server.json

@@ -7,7 +7,7 @@
     "url": "https://github.com/AVIDS2/memorix",
     "source": "github"
   },
-  "version": "1.7.2",
+  "version": "1.8.0",
   "websiteUrl": "https://github.com/AVIDS2/memorix#readme",
   "icons": [
     {
@@ -22,7 +22,7 @@
     {
       "registryType": "npm",
       "identifier": "memorix",
-      "version": "1.7.2",
+      "version": "1.8.0",
       "runtimeHint": "npx",
       "packageArguments": [
         {

package/src/cli/capability-map.ts

@@ -19,6 +19,8 @@ export const CLI_NATIVE_PARITY: Record<string, string> = Object.freeze({
   memorix_transfer: 'memorix transfer export|import',
   memorix_retention: 'memorix retention status|archive|stale',
   memorix_media: 'memorix media import|attach|list|show|generate|status',
+  memorix_evidence: 'memorix evidence list|get|sync|stale|events',
+  memorix_feedback: 'memorix feedback record|show|audit',
   memorix_dashboard: 'memorix dashboard',
   memorix_handoff: 'memorix handoff send',
   memorix_poll: 'memorix poll',