docs(9-ai-sessions): profile memory bank + user-level Bedrock/Vertex + system_prompt_suffix#220
Draft
maximelb wants to merge 3 commits into
Draft
docs(9-ai-sessions): profile memory bank + user-level Bedrock/Vertex + system_prompt_suffix#220maximelb wants to merge 3 commits into
maximelb wants to merge 3 commits into
Conversation
Add a new page explaining the per-profile memory bank that user AI Sessions can use as the user-scoped equivalent of the org-scoped ai_memory hive. Covers mounting semantics, naming rules, limits, REST endpoints, lifecycle (cascade on profile delete, decoupled from session suspend), and the trust model for runner-driven syncs. Cross-links from the existing AI Memory and User Sessions pages, and adds the new endpoints to the API reference. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Document the two remaining features that ship in refractionPOINT/ai-sessions#71 alongside the profile memory bank: 1. User-level Bedrock and Vertex providers — adds Options C and D under "Step 2: Store Claude Credentials" on user-sessions.md, and the two new `POST /v1/auth/claude/bedrock` and `POST /v1/auth/claude/vertex` endpoints on api-reference.md. Updates `GET /v1/auth/claude/status` to show the extended `credential_type` enum (api_key | oauth | bedrock | vertex). Cross-links from alternative-providers.md to the user-side path so readers landing on the org-scoped page can find the user-scoped one. 2. Per-profile system_prompt_suffix — new row in the user-sessions Profile Options table, plus a worked example and the new `system_prompt_suffix_too_long` error response on api-reference.md. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…tion Document the new "## Profile Memory Bank" system-prompt section the runner injects for user-based sessions. Without this hint Claude has no reason to look in /workspace/.memory/ — the Agent SDK has no built-in auto-discovery for arbitrary workspace subdirs. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Documentation for the three features bundled in refractionPOINT/ai-sessions#71.
1. Per-profile memory bank
User-scoped, profile-attached store of markdown files that mounts into the runner workspace at
/workspace/.memory/whenever a session is launched from the profile. The user-side complement to the org-scopedai_memoryHive.2. User-level AWS Bedrock and Google Vertex Claude providers
The same Bedrock / Vertex provider blocks already supported on the org-scoped
ai_agentHive record are now accepted on the per-user BYOK path viaPOST /v1/auth/claude/bedrockandPOST /v1/auth/claude/vertex. Lets a single user route Claude through their own AWS or GCP account from the dashboard, instead of an Anthropic API key or OAuth.3. Per-profile
system_prompt_suffixOptional 16 KB free-form string on a profile that the runner appends to the agent's system prompt whenever a session is launched from that profile. Snapshotted onto the session at creation time so live profile edits don't reach already-running sessions.
Pages added/changed
9-ai-sessions/profile-memory.md— full reference for the memory bank: mounting, sync, naming rules, limits, REST endpoints, lifecycle, security9-ai-sessions/user-sessions.mdsystem_prompt_suffixrow in the Profile Options table9-ai-sessions/api-reference.mdGET/PUT/DELETE /v1/profiles/{id}/memories[/content]endpoint referencePOST /v1/auth/claude/bedrockandPOST /v1/auth/claude/vertexendpoints with request body, response, error responses, and cross-links to provider setupcredential_typeenum onGET /v1/auth/claude/status(api_key | oauth | bedrock | vertex)system_prompt_suffixexample on the Create Profile body and the newsystem_prompt_suffix_too_longerror code9-ai-sessions/memory.md— top-of-page pointer to the per-user equivalent9-ai-sessions/alternative-providers.md— new "Where to configure" subsection cross-linking to the per-user flow on user-sessions.md, so readers landing on the org-scoped page can find the user-scoped onemkdocs.yml— nav entry for the newprofile-memory.mdTest plan
mkdocs build)--data-urlencodefor the path query parameter)system_prompt_suffixmax length matches the implementation (16 KB / 16384 bytes)This is a draft until refractionPOINT/ai-sessions#71 lands and the features ship.
🤖 Generated with Claude Code