memgres 0.5.2__tar.gz → 0.6.0__tar.gz

{memgres-0.5.2 → memgres-0.6.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: memgres
-Version: 0.5.2
+Version: 0.6.0
 Summary: Drop-in memory for AI agents: one Postgres, lexical + semantic recall, diff-versioned history, GDPR-erasable.
 Author: mozgsml
 License-Expression: MIT
@@ -220,6 +220,9 @@ Everything is env, all optional (defaults suit a single-user embed). Full list i
 | `MEMGRES_FULL_BODY` | `false` | force the whole body on every hit (off = auto: short whole, long sliced); `full_body` per call overrides |
 | `MEMGRES_FULL_BODY_MAX_CHARS` | `500` | a body this short is returned whole (`kind="full"`) instead of sliced |
 | `MEMGRES_LIST_PREVIEW_CHARS` | `120` | first-line preview length returned by `memory_list` |
+| `MEMGRES_LIST_BODIES_MAX_BYTES` | `200000` | total body bytes one `bodies=true` browse returns; rows past the cap come back marked `body_omitted`, never dropped |
+| `MEMGRES_MCP_ADMIN_TOOLS` | `auto` | register the `memory_admin_*` control-plane tools over MCP; `auto` = every mode but `single`. A context economy, not a security boundary — each tool authorizes on call |
+| `MEMGRES_MCP_TOOL_VISIBILITY` | `auto` | show each client only the tools its own credential can use, and drop the identity tools where there are no identities (`single`). Display, not authorization — every tool still authorizes on call; `off` lists everything |
 | `MEMGRES_INSTRUCTION` | — | server-side MCP instructions emitted at `initialize` (a client like Claude Code loads it once at connect); unset = omitted; capped at 2 KB |
 | `MEMGRES_VECTOR_BACKEND` | `pgvector` | `pgvector` (same DB) or `qdrant` (set `QDRANT_URL`, `QDRANT_API_KEY`, `MEMGRES_QDRANT_COLLECTION`) |
 | `MEMGRES_EMBED_PROVIDER` | `none` | `none` / `local` / `openai` / `jina` / `openai-compatible` (LM Studio, Ollama, vLLM, TEI…) |
@@ -244,13 +247,31 @@ Everything is env, all optional (defaults suit a single-user embed). Full list i
 | `GET` | `/memories/{id}/blame` | line attribution; `?group`, `?text`, `?lines=1,3-5` |
 | `GET` | `/memories/{id}/at/{seq}` | body reconstructed at a version |
 | `GET` | `/recall` | `?q=&k=&mode=&tags=&path_prefix=&match=&snippet=&full_body=` |
-| `GET` | `/memories` | list a subtree, no query: `?path_prefix=&tags=&limit=&offset=` |
+| `GET` | `/memories` | list a subtree, no query: `?path_prefix=&tags=&limit=&offset=&bodies=` |
+| `GET` | `/whoami` | what this token may do, as capabilities |
+| `GET` | `/admin/orphans` | memories stranded by a `single` → managed switch |
+| `POST` | `/admin/adopt-orphans` | move them into a real namespace (idempotent) |
 | `GET` | `/spaces` | namespaces this token can reach (identity modes) |
+| `POST` | `/spaces` | create one of your own (nothing is created by being named) |
+| `POST` | `/spaces/aliases` | give a reachable namespace a name of your own |
 | `GET` | `/info` | effective config: limits, embed provider/model/dim, recall modes, backend |
 | `GET` | `/healthz` | liveness |
 
+**`{id}` may be a memory's uuid or its tree path** (`/memories/decisions.pricing`)
+— the segment is read as an id when it parses as a uuid, and as a path
+otherwise, so hyphenated and non-ASCII paths (`ops.rate-limits`) address fine.
+A path that a memory has since moved away from is followed on a read
+(the answer sets `moved_from`) and refused on a write, which is what stops an
+edit meant for one memory from quietly becoming a second memory beside it.
+
 Every memory/recall route also takes optional `space` (one of your namespaces by
-name) and `space_id` (canonical id, for shared spaces). In `open`/`managed` mode
+name — your own, one shared with you, or an **alias** you set) and `space_id`
+(canonical id, always unambiguous); a search takes a list of either, or
+`space=all` — every namespace you belong to, plus `space=*` for every namespace
+in the deployment if you are a superadmin — and each hit says which namespace
+answered. Addressing a namespace that does not exist is an error, never a new
+one. In
+`open`/`managed` mode
 the token goes in `Authorization: Bearer <token>` or `X-Memgres-Token`; there are
 also request-access and `/admin/*` provisioning routes — see
 [docs/TENANCY.md](docs/TENANCY.md). OpenAPI/Swagger is at `/docs`. Store errors

{memgres-0.5.2 → memgres-0.6.0}/README.md

@@ -184,6 +184,9 @@ Everything is env, all optional (defaults suit a single-user embed). Full list i
 | `MEMGRES_FULL_BODY` | `false` | force the whole body on every hit (off = auto: short whole, long sliced); `full_body` per call overrides |
 | `MEMGRES_FULL_BODY_MAX_CHARS` | `500` | a body this short is returned whole (`kind="full"`) instead of sliced |
 | `MEMGRES_LIST_PREVIEW_CHARS` | `120` | first-line preview length returned by `memory_list` |
+| `MEMGRES_LIST_BODIES_MAX_BYTES` | `200000` | total body bytes one `bodies=true` browse returns; rows past the cap come back marked `body_omitted`, never dropped |
+| `MEMGRES_MCP_ADMIN_TOOLS` | `auto` | register the `memory_admin_*` control-plane tools over MCP; `auto` = every mode but `single`. A context economy, not a security boundary — each tool authorizes on call |
+| `MEMGRES_MCP_TOOL_VISIBILITY` | `auto` | show each client only the tools its own credential can use, and drop the identity tools where there are no identities (`single`). Display, not authorization — every tool still authorizes on call; `off` lists everything |
 | `MEMGRES_INSTRUCTION` | — | server-side MCP instructions emitted at `initialize` (a client like Claude Code loads it once at connect); unset = omitted; capped at 2 KB |
 | `MEMGRES_VECTOR_BACKEND` | `pgvector` | `pgvector` (same DB) or `qdrant` (set `QDRANT_URL`, `QDRANT_API_KEY`, `MEMGRES_QDRANT_COLLECTION`) |
 | `MEMGRES_EMBED_PROVIDER` | `none` | `none` / `local` / `openai` / `jina` / `openai-compatible` (LM Studio, Ollama, vLLM, TEI…) |
@@ -208,13 +211,31 @@ Everything is env, all optional (defaults suit a single-user embed). Full list i
 | `GET` | `/memories/{id}/blame` | line attribution; `?group`, `?text`, `?lines=1,3-5` |
 | `GET` | `/memories/{id}/at/{seq}` | body reconstructed at a version |
 | `GET` | `/recall` | `?q=&k=&mode=&tags=&path_prefix=&match=&snippet=&full_body=` |
-| `GET` | `/memories` | list a subtree, no query: `?path_prefix=&tags=&limit=&offset=` |
+| `GET` | `/memories` | list a subtree, no query: `?path_prefix=&tags=&limit=&offset=&bodies=` |
+| `GET` | `/whoami` | what this token may do, as capabilities |
+| `GET` | `/admin/orphans` | memories stranded by a `single` → managed switch |
+| `POST` | `/admin/adopt-orphans` | move them into a real namespace (idempotent) |
 | `GET` | `/spaces` | namespaces this token can reach (identity modes) |
+| `POST` | `/spaces` | create one of your own (nothing is created by being named) |
+| `POST` | `/spaces/aliases` | give a reachable namespace a name of your own |
 | `GET` | `/info` | effective config: limits, embed provider/model/dim, recall modes, backend |
 | `GET` | `/healthz` | liveness |
 
+**`{id}` may be a memory's uuid or its tree path** (`/memories/decisions.pricing`)
+— the segment is read as an id when it parses as a uuid, and as a path
+otherwise, so hyphenated and non-ASCII paths (`ops.rate-limits`) address fine.
+A path that a memory has since moved away from is followed on a read
+(the answer sets `moved_from`) and refused on a write, which is what stops an
+edit meant for one memory from quietly becoming a second memory beside it.
+
 Every memory/recall route also takes optional `space` (one of your namespaces by
-name) and `space_id` (canonical id, for shared spaces). In `open`/`managed` mode
+name — your own, one shared with you, or an **alias** you set) and `space_id`
+(canonical id, always unambiguous); a search takes a list of either, or
+`space=all` — every namespace you belong to, plus `space=*` for every namespace
+in the deployment if you are a superadmin — and each hit says which namespace
+answered. Addressing a namespace that does not exist is an error, never a new
+one. In
+`open`/`managed` mode
 the token goes in `Authorization: Bearer <token>` or `X-Memgres-Token`; there are
 also request-access and `/admin/*` provisioning routes — see
 [docs/TENANCY.md](docs/TENANCY.md). OpenAPI/Swagger is at `/docs`. Store errors

{memgres-0.5.2 → memgres-0.6.0}/memgres/__init__.py

@@ -27,7 +27,8 @@ from .identity import (
     Principal, AuthError, SpaceNotFound,
     resolve, resolve_space, new_token, valid_format,
     create_user, create_namespace, list_spaces,
-    issue_token, register_token, revoke_token, list_tokens,
+    list_users, list_namespaces, list_members,
+    issue_token, register_token, revoke_token, list_tokens, token_owner,
     request_access, approve_request, deny_request, list_requests,
 )
 
@@ -43,6 +44,7 @@ __all__ = [
     "Principal", "AuthError", "SpaceNotFound",
     "resolve", "resolve_space", "new_token", "valid_format",
     "create_user", "create_namespace", "list_spaces",
-    "issue_token", "register_token", "revoke_token", "list_tokens",
+    "list_users", "list_namespaces", "list_members",
+    "issue_token", "register_token", "revoke_token", "list_tokens", "token_owner",
     "request_access", "approve_request", "deny_request", "list_requests",
 ]

{memgres-0.5.2 → memgres-0.6.0}/memgres/_version.py

@@ -8,4 +8,4 @@ here at release; nowhere else carries the number.
 PEP 440: a ``.devN`` suffix marks an unreleased build ahead of the last tag.
 """
 
-__version__ = "0.5.2"
+__version__ = "0.6.0"