hack-ink · yvette-carlisle · Jun 8, 2026 · Jun 8, 2026
diff --git a/README.md b/README.md
@@ -116,26 +116,26 @@ flowchart TB
 Quick comparison snapshot (objective/high-level).
 This table compares capability coverage, not overall project quality.
 
-| Capability | ELF | OpenViking | mem0 | qmd | claude-mem | memsearch |
-| ---------- | --- | ---------- | ---- | --- | ---------- | --------- |
-| Local-first self-hosted workflow | ✅ | ✅ | ✅ (OpenMemory) | ✅ | ✅ | ✅ |
-| MCP integration | ✅ | — | ✅ (OpenMemory) | ✅ | ✅ | ⚠️ |
-| CLI-first developer workflow | — | ✅ | — | ✅ | ⚠️ | ✅ |
-| HTTP API service surface | ✅ | ✅ | ✅ | ⚠️ (MCP Streamable HTTP) | ✅ | — |
-| Query expansion or query rewriting | ✅ | ✅ | ⚠️ | ✅ | — | — |
-| LLM reranking stage | ✅ | ⚠️ | ⚠️ | ✅ | — | — |
-| Hybrid dense + sparse retrieval | ✅ | ✅ | ⚠️ | ✅ | ✅ | ✅ |
-| Progressive disclosure style retrieval | ✅ | ✅ | — | — | ✅ | ⚠️ |
-| Evidence-bound memory writes | ✅ | — | — | — | — | — |
-| Deterministic and LLM-ingestion boundary | ✅ | ⚠️ | ⚠️ | — | — | — |
-| Source-of-truth + rebuildable derived index | ✅ | ✅ | ⚠️ | ⚠️ | ⚠️ | ✅ |
-| Hierarchical/recursive retrieval strategy | ⚠️ (in progress) | ✅ | ⚠️ | ⚠️ | ⚠️ | ⚠️ |
-| Progressive context loading (L0/L1/L2 style) | ⚠️ (in progress) | ✅ | ⚠️ | — | ⚠️ | — |
-| Built-in web memory inspector/viewer | — | — | ✅ (OpenMemory) | — | ✅ | — |
-| Hosted managed option | — | — | ✅ | — | — | — |
-| Multi-tenant scope semantics | ✅ | ⚠️ | ✅ | — | — | — |
-| TTL/lifecycle policy controls | ✅ | ⚠️ | ✅ | — | ⚠️ | — |
-| Graph memory mode | ⚠️ (graph-lite: structured relations persisted; optional search `relation_context`) | ⚠️ (URI-link relations) | ✅ (optional) | — | — | — |
+| Capability | ELF | agentmemory | OpenViking | mem0 | qmd | claude-mem | memsearch |
+| ---------- | --- | ----------- | ---------- | ---- | --- | ---------- | --------- |
+| Local-first self-hosted workflow | ✅ | ✅ | ✅ | ✅ (OpenMemory) | ✅ | ✅ | ✅ |
+| MCP integration | ✅ | ✅ | — | ✅ (OpenMemory) | ✅ | ✅ | ⚠️ |
+| CLI-first developer workflow | — | ✅ | ✅ | — | ✅ | ⚠️ | ✅ |
+| HTTP API service surface | ✅ | ✅ | ✅ | ✅ | ⚠️ (MCP Streamable HTTP) | ✅ | — |
+| Query expansion or query rewriting | ✅ | ⚠️ | ✅ | ⚠️ | ✅ | — | — |
+| LLM reranking stage | ✅ | ⚠️ | ⚠️ | ⚠️ | ✅ | — | — |
+| Hybrid dense + sparse retrieval | ✅ | ✅ | ✅ | ⚠️ | ✅ | ✅ | ✅ |
+| Progressive disclosure style retrieval | ✅ | ⚠️ | ✅ | — | — | ✅ | ⚠️ |
+| Evidence-bound memory writes | ✅ | — | — | — | — | — | — |
+| Deterministic and LLM-ingestion boundary | ✅ | ⚠️ | ⚠️ | ⚠️ | — | — | — |
+| Source-of-truth + rebuildable derived index | ✅ | ⚠️ | ✅ | ⚠️ | ⚠️ | ⚠️ | ✅ |
+| Hierarchical/recursive retrieval strategy | ⚠️ (in progress) | ⚠️ | ✅ | ⚠️ | ⚠️ | ⚠️ | ⚠️ |
+| Progressive context loading (L0/L1/L2 style) | ⚠️ (in progress) | ⚠️ | ✅ | ⚠️ | — | ⚠️ | — |
+| Built-in web memory inspector/viewer | — | ✅ | — | ✅ (OpenMemory) | — | ✅ | — |
+| Hosted managed option | — | — | — | ✅ | — | — | — |
+| Multi-tenant scope semantics | ✅ | ⚠️ | ⚠️ | ✅ | — | — | — |
+| TTL/lifecycle policy controls | ✅ | ⚠️ | ⚠️ | ✅ | — | ⚠️ | — |
+| Graph memory mode | ⚠️ (graph-lite: structured relations persisted; optional search `relation_context`) | ⚠️ | ⚠️ (URI-link relations) | ✅ (optional) | — | — | — |
 
 Legend: `✅` built-in and documented; `⚠️` partial, optional, or in-progress; `—` not a first-class documented capability.
 
@@ -144,6 +144,7 @@ Project signature strengths (what each does especially well):
 | Project | Signature strengths | Potential ELF adoption value |
 | ------- | ------------------- | ---------------------------- |
 | ELF | Evidence-bound writes, deterministic ingestion boundary, SoT + rebuildable index, eval tooling | Keep as core differentiators while extending retrieval and UX |
+| agentmemory | Cross-agent hooks, MCP/REST packaging, local viewer, iii console observability, coding-agent continuity benchmarks | Use as adapter/baseline and UX reference, not a replacement for ELF provenance semantics |
 | OpenViking | Filesystem-like context model (`viking://`), hierarchical retrieval, staged retrieval trajectory | Improve query planning, recursive retrieval, and explainable stage outputs |
 | mem0 | Broad ecosystem (SDK + hosted + OpenMemory), multi-entity scope, lifecycle + optional graph memory | Strengthen event/history APIs and additive graph context channel |
 | qmd | High-quality local retrieval pipeline (query expansion + weighted fusion + rerank), strong CLI/MCP workflow | Borrow transparent routing/fusion knobs and local debugging ergonomics |
@@ -154,8 +155,9 @@ Detailed comparison, mechanism-level analysis, and source map:
 
 - [Detailed External Comparison](docs/guide/research/comparison_external_projects.md)
 - [Research Projects Inventory](docs/guide/research/research_projects_inventory.md)
+- [Agent Memory Selection Research Run](docs/research/2026-06-08-agent-memory-selection.json)
 
-Snapshot date in that document: February 17, 2026.
+Latest external research refresh: June 8, 2026.
 
 ## Documentation
 

diff --git a/docs/governance.md b/docs/governance.md
@@ -24,6 +24,7 @@ The split between `spec` and `guide` is by task shape, not by reader type.
 | --- | --- | --- | --- | --- |
 | Spec | `docs/spec/` | What must be true? | Contracts, schemas, invariants, required behavior | Any behavior or schema change |
 | Guide | `docs/guide/` | What should I do? | Runbooks, migrations, validation, troubleshooting | Any procedure or operational change |
+| Research runs | `docs/research/` | Which evidence-backed research run reached what state? | Machine-readable hypotheses, evidence, trade-offs, challenge records, and terminal decision state | A research workflow needs durable replayable state |
 | Plan artifacts | `docs/plans/` | Which saved plan artifact should a planning tool or execution workflow use? | Tool-managed planning outputs | As emitted or updated by the relevant tool |
 
 ## Placement rules
@@ -32,6 +33,8 @@ The split between `spec` and `guide` is by task shape, not by reader type.
 - If a document defines actions, it belongs in `docs/guide/`.
 - If a document is non-normative decision support, comparison, or research input, treat it
   as guide-class material and store it under `docs/guide/`.
+- If a research workflow requires a machine-readable run file with replayable events,
+  store that run file under `docs/research/` and link to it from the relevant guide.
 - Do not treat `docs/plans/` as a general-purpose docs bucket.
 - Use `docs/plans/` only for artifacts produced or consumed by planning tools or
   workflows that explicitly depend on saved plan files.
@@ -85,7 +88,9 @@ When answering a repository question:
    - "What must be true?" -> `docs/spec/index.md`
    - "What should I do?" -> `docs/guide/index.md`
 3. Read `Makefile.toml` when the task depends on repository automation or named tasks.
-4. Use `docs/plans/` only when the task explicitly concerns a saved plan artifact used by
+4. Use `docs/research/` only when the task explicitly concerns a machine-readable
+   research run file used by a research workflow.
+5. Use `docs/plans/` only when the task explicitly concerns a saved plan artifact used by
    a planning tool or execution workflow.
 
 ## Update workflow

diff --git a/docs/guide/research/comparison_external_projects.md b/docs/guide/research/comparison_external_projects.md
@@ -8,6 +8,8 @@ Outputs: A comparison matrix and trade-off summary suitable for follow-up design
 
 Scope note: This document is intentionally detailed and source-heavy. Keep `README.md` concise and link here for full analysis.
 For a full list of reviewed and pending projects, see `docs/guide/research/research_projects_inventory.md`.
+For the June 2026 agentmemory and dreaming decision run, see
+`docs/research/2026-06-08-agent-memory-selection.json`.
 
 Comparison focuses on shared capabilities, ELF distinctives, and objective trade-offs. These projects solve adjacent problems, but their primary storage units and default workflows differ.
 
@@ -30,6 +32,91 @@ Legend:
 Note: In this section, mem0 refers to the Mem0 ecosystem, including OpenMemory (an MCP memory server with a built-in UI).
 OpenViking is included as a newly reviewed project with mechanism-level analysis.
 
+## June 2026 Agentmemory And Dreaming Refresh
+
+Snapshot date for this subsection: June 8, 2026.
+
+This refresh re-evaluates ELF after the June 2026 hardening work and after the
+appearance of [agentmemory](https://github.com/rohitg00/agentmemory) as a high-velocity
+coding-agent memory project. It also records the current vendor direction around
+dreaming-style background memory consolidation.
+
+### Current ELF Position
+
+ELF remains strongest as a high-trust memory service rather than a turnkey coding-agent
+continuity plugin. The current main branch has:
+
+- evidence-linked fact writes and quote-bound provenance;
+- deterministic `add_note` separated from LLM-driven `add_event`;
+- Postgres as source of truth and Qdrant as a rebuildable derived index;
+- scoped HTTP/MCP service semantics, TTL/lifecycle policy, graph-lite relation context,
+  and retrieval evaluation tooling;
+- recently restored local gates, stricter config presence, generated OpenAPI/Scalar docs,
+  and Docker Compose service dependencies.
+
+### agentmemory
+
+agentmemory is now important enough to track as a first-class comparison target. Its
+public README advertises cross-agent support for Claude Code, Codex CLI, Cursor, Gemini
+CLI, OpenCode, and generic MCP clients; MCP/REST access; hook-based capture; hybrid
+BM25/vector/graph retrieval; consolidation/lifecycle behavior; a local viewer on `:3113`;
+and iii console observability for traces, KV state, triggers, queues, and streams. Its
+roadmap still lists benchmark CI, session replay UI, governance baseline, enterprise trust
+features, and a v1.0 stability freeze as future work.
+
+ELF implication: do not replace ELF with agentmemory. Treat it as:
+
+- an optional capture/import adapter for coding-agent session observations;
+- a benchmark and UX baseline for local continuity workflows;
+- a source of product ideas around hooks, viewer, replay, audit, and tool breadth.
+
+### Dreaming And Background Consolidation
+
+OpenAI frames dreaming as background curation that synthesizes memory state, applies
+preferences, and keeps memory current over time. Anthropic Claude Dreams is the strongest
+safety reference: a dream reads an input memory store plus 1-100 sessions, produces a
+separate output memory store, never modifies the input store, and leaves the output
+reviewable, attachable, discardable, archivable, or deletable. Google examples add two
+operator patterns: Always-On Memory Agent runs scheduled consolidation, while Gemini CLI
+Auto Memory mines idle transcripts but writes reviewable patches and skill drafts to an
+inbox before anything is applied.
+
+ELF implication: dreaming should be a reviewed derived layer over authoritative evidence,
+not a destructive rewrite path. The target shape is:
+
+- immutable observations, notes, events, traces, and source pointers as input;
+- asynchronous consolidation jobs that produce candidate derived memories, pages, graph
+  views, or skills;
+- explicit lineage, diff, confidence, contradiction/staleness markers, and review/apply
+  controls;
+- rebuildable outputs that can be discarded without corrupting source-of-truth memory.
+
+### Current Recommendation
+
+Continue building ELF. Do not directly adopt agentmemory or managed dreaming as the core
+backend. The next work should prioritize:
+
+1. a reviewable derived consolidation pipeline;
+2. read-only viewer plus retrieval/consolidation observability;
+3. optional agentmemory import/baseline adapter;
+4. graph-lite typed query and derived knowledge pages with provenance/lint.
+
+This ordering reuses the existing vNext planning surface instead of starting a parallel
+roadmap: [XY-286](https://linear.app/hack-ink/issue/XY-286/knowledge-memory-derived-entityconceptproject-pages-with-provenance),
+[XY-19](https://linear.app/hack-ink/issue/XY-19/add-a-read-only-web-viewer-for-sessions-and-traces),
+[XY-27](https://linear.app/hack-ink/issue/XY-27/viewer-add-retrieval-observability-panels-on-top-of-the-read-only),
+and [XY-70](https://linear.app/hack-ink/issue/XY-70/graph-lite-dx-typed-schema-typed-query-nanograph-inspired)
+remain the right backbone.
+
+Primary sources for this refresh:
+
+- https://github.com/rohitg00/agentmemory
+- https://raw.githubusercontent.com/rohitg00/agentmemory/main/ROADMAP.md
+- https://openai.com/index/chatgpt-memory-dreaming/
+- https://platform.claude.com/docs/en/managed-agents/dreams
+- https://github.com/GoogleCloudPlatform/generative-ai/tree/main/gemini/agents/always-on-memory-agent
+- https://github.com/google-gemini/gemini-cli/blob/main/docs/cli/auto-memory.md
+
 ## Scope And Intended Use
 
 | Aspect             | ELF                                                   | [memsearch](https://github.com/zilliztech/memsearch) | [qmd](https://github.com/tobi/qmd) | [claude-mem](https://github.com/thedotmack/claude-mem) | [mem0](https://github.com/mem0ai/mem0) |

diff --git a/docs/guide/research/index.md b/docs/guide/research/index.md
@@ -0,0 +1,18 @@
+# Research Guide Index
+
+Goal: Route agents to external comparison and decision-support research for ELF memory architecture.
+Read this when: You need to compare ELF with adjacent memory, context, RAG, or consolidation systems.
+Inputs: Current ELF docs/code, public external project docs, tracker state, and checked-in research run files.
+Depends on: `docs/index.md`, `docs/governance.md`, and `docs/research/` for machine-readable research runs.
+Outputs: The smallest comparison or inventory document needed for implementation decisions.
+
+## Documents
+
+- `research_projects_inventory.md`: audited and pending external projects, research depth, and current planning surface.
+- `comparison_external_projects.md`: detailed capability comparison, project trade-offs, source map, and research-backed ELF directions.
+
+## Machine-Readable Runs
+
+Machine-authoritative research run JSON files live under `docs/research/`.
+Use those files when a research conclusion needs replayable hypotheses, evidence,
+trade-offs, challenge records, and terminal decision state.
diff --git a/docs/guide/research/research_projects_inventory.md b/docs/guide/research/research_projects_inventory.md
@@ -6,7 +6,7 @@ Inputs: Existing research notes, open architecture questions, and tracked adopti
 Depends on: `docs/guide/research/comparison_external_projects.md`.
 Outputs: A current inventory of reviewed and pending external projects.
 
-Last updated: April 17, 2026.
+Last updated: June 8, 2026.
 
 ## Legend
 
@@ -18,6 +18,10 @@ Last updated: April 17, 2026.
 
 | Project | Research depth | Current status | Why it matters to ELF | Primary reference |
 | ------- | -------------- | -------------- | --------------------- | ----------------- |
+| [agentmemory](https://github.com/rohitg00/agentmemory) | D1 | Reviewed | Cross-agent coding-memory hooks, MCP/REST surface, viewer, consolidation lifecycle, and external benchmark target | `docs/guide/research/comparison_external_projects.md`; `docs/research/2026-06-08-agent-memory-selection.json` |
+| [OpenAI ChatGPT Memory Dreaming](https://openai.com/index/chatgpt-memory-dreaming/) | D1 | Reviewed | Background memory synthesis and staleness repair as a product direction | `docs/research/2026-06-08-agent-memory-selection.json` |
+| [Claude Managed Agents Dreams](https://platform.claude.com/docs/en/managed-agents/dreams) | D1 | Reviewed | Reviewable derived memory-store output over past sessions; strong safety shape for ELF consolidation | `docs/research/2026-06-08-agent-memory-selection.json` |
+| [Gemini CLI Auto Memory](https://github.com/google-gemini/gemini-cli/blob/main/docs/cli/auto-memory.md) | D1 | Reviewed | Background session mining with project-local review inbox for memory patches and skills | `docs/research/2026-06-08-agent-memory-selection.json` |
 | [mem0](https://github.com/mem0ai/mem0) | D2 | Reviewed | Graph memory as additive context, memory history and async mode trade-offs | `docs/guide/research/comparison_external_projects.md` |
 | [memsearch](https://github.com/zilliztech/memsearch) | D2 | Reviewed | Markdown-first SoT + rebuildable index pattern | `docs/guide/research/comparison_external_projects.md` |
 | [qmd](https://github.com/tobi/qmd) | D2 | Reviewed | Retrieval routing, weighted fusion, and local-first explainability | `docs/guide/research/comparison_external_projects.md` |
@@ -35,6 +39,26 @@ Last updated: April 17, 2026.
 | [LightRAG](https://github.com/HKUDS/LightRAG) | D0 | Pending deep dive | Graph-augmented RAG strategy relevance; not yet audited to adoption level | Discussion history only |
 | [GraphRAG](https://www.microsoft.com/en-us/research/project/graphrag/) | D0 | Pending deep dive | Graph-based retrieval concepts; not yet audited to implementation decision level | Discussion history only |
 
+## June 2026 Activity Snapshot
+
+GitHub API snapshot time: 2026-06-08T06:01:57Z.
+
+The monitored project set is still moving quickly. Recent push activity was observed for
+agentmemory, mem0, qmd, claude-mem, OpenViking, gbrain, graphify, LangGraph, Graphiti,
+RAGFlow, LightRAG, and GraphRAG. Notable current scale signals:
+
+- agentmemory: 21,783 stars, latest release `v0.9.27`, pushed 2026-06-07.
+- mem0: 58,005 stars, latest release `cli-node-v0.2.8`, pushed 2026-06-06.
+- claude-mem: 81,157 stars, latest release `v13.4.1`, pushed 2026-06-08.
+- graphify: 62,294 stars, latest release `v0.8.35`, pushed 2026-06-07.
+- RAGFlow: 82,150 stars, latest release `v0.25.6`, pushed 2026-06-08.
+- LightRAG: 36,270 stars, latest release `v1.5.0`, pushed 2026-06-08.
+- GraphRAG: 33,545 stars, latest release `v3.1.0`, pushed 2026-06-05.
+
+Interpretation: this is not a settled market. ELF should keep watching external
+implementation velocity, but the current activity signal alone does not justify
+replacing ELF's evidence-bound service contract.
+
 ## Current Planning Surface
 
 - Linear project: [ELF vNext: Evidence-to-Knowledge Memory](https://linear.app/hack-ink/project/elf-vnext-evidence-to-knowledge-memory-d7a9dd3f3e86)
@@ -46,6 +70,8 @@ Last updated: April 17, 2026.
   - [XY-40](https://linear.app/hack-ink/issue/XY-40/vision-track-elf-as-a-high-trust-memory-system-for-singlemulti-agent)
   - [XY-51](https://linear.app/hack-ink/issue/XY-51/agent-memory-ux-mcp-surface-skills-doc-pointers-epic)
   - [XY-63](https://linear.app/hack-ink/issue/XY-63/research-openviking-as-optional-doc-backend-integration-sketch)
+- Current June 2026 research run:
+  - `docs/research/2026-06-08-agent-memory-selection.json`
 
 ## Notes