Skip to content

Solution for long term memory for agent coding CLIs and to facilitate handoff between different agent vendors

Rust

Merge env-file BOM fix from main

AkitaOnRails552e8d9993ba
diff --cc CHANGELOG.md
index 5a20d98f,63b5437f..fbda7e1c
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@@ -11,180 -11,10 +11,183 @@@ version in [`docs/changelog/`](docs/cha
## [Unreleased]
+### Added
+
+- Added a first-party Hermes Agent lifecycle and an `install-mcp --client
+ hermes` client. The `hooks:` block `install-hooks --agent hermes` prints now
+ also wires `on_session_start`, `pre_llm_call` (the prompt, never the
+ conversation history), `on_session_end` as a turn boundary and
+ `on_session_finalize` as the session end, and each primary-session
+ `pre_llm_call` checks for a pending handoff and brings it into that turn as
+ `{"context": ...}`. This includes a resumed native Hermes session. Hermes
+ events route by `TERMINAL_CWD` when it names an existing directory (in a
+ gateway, for the default profile only), and a delegated child's lifecycle
+ events are dropped. When the
+ community memory-provider plugin is installed, only the tool events are
+ printed. `install-mcp --client hermes` prints an `mcp_servers:` entry for the
+ session-less stdio bridge with no token in it, and refuses `--apply` ([#1454](https://github.com/akitaonrails/ai-memory/pull/1454)).
+- Added bounded local document extraction to the optional
+ `ai-memory-importer` companion. Its dry-run-first `document` command accepts
+ UTF-8 text/Markdown, text-layer PDF, DOCX, PPTX, XLSX, VTT, and
+ SRT; emits attributed, client-sanitized markdown; caps source, archive,
+ extracted text, PDF streams, sheets, and actual XLSX cells; and uses deterministic digest paths plus
+ conditional MCP writes for idempotent reruns without overwriting an existing
+ foreign page. OCR, images, and raw audio/video remain explicitly out of
+ scope.
+- Added passive per-workload LLM telemetry to `ai-memory status --json` and
+ `GET /admin/status`. The closed workload set covers consolidation, reranking,
+ auto-improvement, bootstrap, query answers, lint, profile maintenance, dream
+ merges, and experience abstraction, with process-lifetime calls, outcomes,
+ latency buckets, coarse token estimates, and provider-reported usage where available.
+ It records no prompt or output text and does not change provider/model
+ routing; public per-workload profiles remain deferred until measurements
+ justify their configuration and authentication surface.
+- Added structured pre-rank filters to `memory_query`: case-insensitive
+ `kind` and exact-array `tag`, typed `tier`, and literal `path_prefix`.
+ Filters apply consistently to project/scopes hybrid streams, cross-project
+ global FTS, `as_of` history, supplemental profile hits, and `pin_first`.
+ Filtered misses omit raw observations because those rows carry no page
+ metadata.
+- Added a read-only managed-workstream operations page at
+ `/web/workstreams`. On multi-user servers it is root-only; single-operator
+ installs retain their existing admin access. The bounded overview orders
+ live and interrupted work first and derives `active`, `interrupted`,
+ `completed`, and `empty` only from trusted run lifecycle fields and imported
+ event counts. It exposes no checkout paths, repository fingerprints, lease
+ owners, native session ids, source metadata, or transcript content, and it
+ does not broaden checkout-local `resume` discovery.
+- Added `ai-memory recent` and `GET /admin/recent` to list the most recently
+ updated wiki pages, newest first. The listing is scoped to the current
+ project by default (cwd / marker, or `--workspace`/`--project`); `--global`
+ covers every project and labels each hit with its workspace and project.
+ Each hit carries an RFC 3339 `updated_at` instead of an FTS rank. The admin
+ route is root-only on a multi-user server, the same as `/admin/search`.
+ ([#1379](https://github.com/akitaonrails/ai-memory/pull/1379))
+- Added an explicit `serve --allow-insecure-human-auth` (or
+ `AI_MEMORY_ALLOW_INSECURE_HUMAN_AUTH=true`) escape hatch for operators who
+ knowingly run password login over plain HTTP on a private LAN.
+ It is off by default, does not bypass the separate unauthenticated-exposure
+ guard, and prints an unconditional warning that passwords, API credentials,
+ session cookies, and memory data cross the network unencrypted. (#1383)
+- Added first-party Cline CLI lifecycle and MCP support. `install-hooks
+ --agent cline --apply` writes Cline's nine supported event files globally or
+ per project, refuses user-owned and symlinked targets, and keeps embedded
+ bearer credentials owner-only. `install-mcp --client cline --apply` writes
+ Cline's top-level transport fields and explicitly selects Streamable HTTP.
+ Detached start hooks recover handoffs through MCP; yolo mode disables hooks.
+ Also added `install-mcp --client freebuff --apply`, which registers the
+ sessionless stdio bridge in `~/.agents/mcp.json` without starting another
+ ai-memory server. (#1335)
+- Added `[ingest] drop_events` (env `AI_MEMORY_INGEST__DROP_EVENTS`), an
+ opt-in list of hook event kinds the server drops before storing them, next
+ to the existing `drop_tool_use_id_prefixes` and `drop_tool_names`. A dropped
+ event gets the same 202, `dropped_by_policy` metric and log as the other
+ policy drops, before the rate limiter, the ingest semaphore and the
+ sanitizer. Subagent classification precedes every id, name, and kind matcher,
+ so a matching marked event still seeds tracking for its unmarked tail. The
+ default drops nothing. `Config::load` checks each value
+ against the typed event kind and rejects unknown kinds and `session-start`
+ and `session-end`, which open and close sessions. The motivating case is
+ `pre-tool-use`: in one production store about 87% of those observations had
+ an empty body, and each pairs with a `post-tool-use` that carries the
+ content. With `pre-tool-use` dropped, a session whose only work was a tool
+ call that never finished now ends as lifecycle-only. (#1339)
+- Added opt-in `[sessions] idle_end_after_hours` (env
+ `AI_MEMORY_SESSIONS__IDLE_END_AFTER_HOURS`; default `0`, off; range
+ `1..=8760`). When it is set, the `serve` maintenance tick ends open sessions
+ with no activity for that many hours, such as a harness that crashed or a
+ terminal closed before `SessionEnd` fired. A substantive session gets its
+ `sessions/<id>.md` summary page, attributed to the session's owner, and its
+ consolidation job. A boundary-only session ends lifecycle-only and releases
+ the handoff it had accepted. The tick creates no handoff of its own and skips
+ sessions bound to an active managed run. A real `SessionEnd` or checkpoint
+ that lands while the tick runs wins, and the tick leaves that session alone.
+ `ended_at` is the time of the tick; the session's last activity goes in the
+ audit row. (#1338)
+- Added first-party DeepSeek Harness (DSH) lifecycle capture. `install-hooks
+ --agent dsh` (aliases `deepseek-harness`, `deepseek_harness`) stages the
+ new `hooks/dsh` bundle, writes a generated Claude Code `hooks.json`, and
+ prints a Cordis patch fragment that mounts DSH's official
+ `@deepseek-ai/dsh-hooks-claude-code` bridge on
+ `$DSH_HOME/profiles/<profile>/cordis.patch.yml` (the patch file itself is
+ never auto-written). DSH sessions are stored as `agent=dsh`, and its bridge
+ forwards Claude Code-shaped tool payloads, so `[capture] ignore_paths`
+ exclusions apply to its file, search and shell calls (Windows `pwsh`, POSIX
+ `bash`) like any other Claude-shaped adapter; the reserved Code Mode
+ `run_code` transport that its `headless` profile calls is treated as inline
+ program source. The bridge
+ delivers exactly seven events (no `PreCompact` or `SessionEnd`) and runs
+ `SessionStart` detached, so finished sessions close with
+ `finalize-session --agent dsh` and pending handoffs are recovered via
+ `memory_handoff_list` / `memory_handoff_accept`. (#1337)
+- Added an Antigravity skill root family. `install-skills --agent antigravity`
+ (aliases `antigravity-cli`, `agy`; also `install-instructions
+ --skills-agent antigravity`) writes the managed Agent Skills to both global
+ roots Antigravity reads, `~/.gemini/config/skills` (App and IDE) and
+ `~/.gemini/antigravity-cli/skills` (CLI), or to the workspace
+ `.agents/skills` with `--scope project`. `ai-memory setup` now uses it for
+ Antigravity, which until now got its skills only in `~/.agents/skills`, a
+ directory Antigravity does not read globally, and `uninstall --only skills`
+ sweeps both roots. MCP-driven routing refresh and `backup-agents` expose and
+ preserve the same roots, including the shared project `.agents/skills`.
+ (#1357)
+
+- Added a read-only wiki link graph in `/web`. The header **Graph** link opens
+ `/web/graph`, a server-rendered map of resolved wikilinks whose endpoints sit
+ in different projects, plus a table of those edges. Each project page links
+ to `/web/w/<workspace>/<project>/graph`, which shows that project's latest
+ unexpired pages and the resolved same-project links between them as an SVG
+ and a table. Both pages are HTML with no JavaScript.
+ `GET /api/v1/graph` now accepts `workspace` and `project` together to return
+ only edges that touch that project (passing only one of them is `400`), and
+ `GET /api/v1/workspaces/{workspace}/projects/{project}/graph` returns the
+ in-project node and edge lists. A project graph keeps at most 120 pages,
+ preferring those with the most same-project neighbours and then the most
+ recently updated; the store applies that cap before loading nodes or their
+ selected edges into the process. Expired pages stay hidden. An edge still
+ requires both ends to be readable.
+- Added Antigravity hook parity with Claude Code and Codex for the `agy` CLI,
+ the desktop App and the IDE. `install-hooks --agent antigravity-cli` now
+ wires the native `SessionStart`, which injects the handoff as an
+ `injectSteps` system message that stays in the conversation, where the
+ ephemeral message it replaces was dropped after the first turn.
+ `PreInvocation` gets its own `pre-invocation` handler: on builds without
+ `SessionStart`, the first turn's call stands in for it, once per
+ conversation. The native hook now reads each turn's prompts from the
+ conversation's own transcript (`transcriptPath`, only under
+ `~/.gemini/<app>/brain/<conversationId>/`, no symlinks out of it, 8 MiB per
+ hook and a 100 ms filesystem deadline) and records them as user prompts,
+ so summaries and handoffs carry them. A timed-out read leaves its cursor
+ untouched and retries on the next hook. A `Stop` that finished the turn
+ (`NO_TOOL_CALL` or `model_stop`) with
+ `fullyIdle: true` writes a turn checkpoint, the path OpenCode 2 uses: the
+ summary page and one handoff are refreshed and the conversation stays open,
+ so `finalize-session` is no longer needed for normal use. It remains the
+ explicit close, hands the baton over without the 10-minute quiet period,
+ queues opt-in consolidation, and skips its `history.jsonl` prompt replay
+ for a conversation whose prompts were already captured. Rerun
+ `install-hooks --agent antigravity-cli --apply` to add `SessionStart`;
+ existing native-hook installs get the prompts and checkpoints without it.
+ A turn checkpoint never ends a conversation, even a turn with no captured
+ prompt or tool call, so a handoff it claimed at start is never offered
+ again; the conversation start is claimed atomically, so a concurrent
+ `SessionStart` and first `PreInvocation` cannot both claim a handoff;
+ prompt ingest keys include the prompt's time and text, so a rewind that
+ reuses step numbers keeps its new prompts; and the transcript is opened
+ with `O_NOFOLLOW | O_NONBLOCK` and, on Linux, checked against the opened
+ descriptor's real path. (#1416)
+
### Fixed
+- Reserved local-embedding capacity for interactive search queries instead of
+ allowing concurrent page writes and backfill work to occupy every inference
+ slot. Excess queries fall back immediately to non-vector retrieval, and a
+ cancelled request retains its slot until its non-cancellable CPU inference
+ actually ends. An opt-in real-model contention probe records idle,
+ unbounded-load, and bounded-load p50/p95 query latency; hosted embedding
+ providers are unchanged.
+ - Ignored a UTF-8 byte-order mark at the start of `run --env-file` input so
+ the first environment variable keeps its intended name.
+ ([#1465](https://github.com/akitaonrails/ai-memory/pull/1465))
- Invoked GitHub Copilot CLI's native Windows lifecycle commands with
PowerShell's call operator. Quoted `ai-memory.exe` paths now execute instead
of producing a parser error that made fail-closed pre-tool hooks deny every