Repository navigation
add: Context Graph user docs under AI memory - #1847
Merged
katarinasupe merged 15 commits intoOct 23, 2026
Merged
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
katarinasupe
self-requested a review
October 21, 2026 09:10
…ply the docs rules
…config set for all keys, memory hand-back, setup's own Memgraph
…ies, hook trust for Codex and Grok, doctor output, empty capture.connectors, entity merging, SDK plugin
antejavor
commented
Oct 22, 2026
antejavor
added a commit
to memgraph/ai-toolkit
that referenced
this pull request
Oct 22, 2026
… anthropic 2.x Found by a second audit of the AI memory docs (memgraph/documentation#1847). - setup: a local --memgraph-url that doesn't answer (down, or refusing the login because credentials weren't set yet) is no longer swapped for setup's own container on another port; setup stops and names both ports. - sessions-graph derive: structured outputs through messages.create's output_config, which anthropic>=0.77 takes; 1.0 dropped the beta's output_format, so every candidate failed on a fresh install. derive now exits 1 when every candidate errored. - ContextGraphMemory.end_session's docstring says what ending does: embed what's left and mark the session for reconciliation.
…ntology learning in the diagram; runnable SDK examples, credentials before setup, review fixes
…logy as one term, full recall and extraction key names, memory.backend default, why there are two install commands, section intros and shorter paragraphs
antejavor
commented
Oct 23, 2026
…o The graph, recall and extraction settings and setup flags on Configuration, Supported agents and Custom agents pages, clearer Get started steps
antejavor
marked this pull request as ready for review
October 23, 2026 09:54
antejavor
requested review from
Josipmrden,
gitbuda and
vpavicic
as code owners
October 23, 2026 09:54
katarinasupe
merged commit Oct 23, 2026
9a2c8f4
into
update/ai-ecosystem-structure-v2
5 checks passed
katarinasupe
added a commit
that referenced
this pull request
Oct 23, 2026
* update: Reorganize the docs sidebar - Move AI ecosystem above Client libraries. - Move Use cases under AI ecosystem. - Move Clustering under Database management. - Move Data streams under Data migration. - Move Memgraph Lab next to Memgraph Zero and set both apart with sidebar dividers. Moved pages keep working at their old URLs through permanent redirects, including the .md copies listed in llms.txt. Internal links and existing redirects now point at the new URLs directly. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix: Keep the original CRLF line endings in edited pages Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * update: Restructure the AI ecosystem section - Fold GraphRAG Prerequisites into the GraphRAG overview. - Move SQL2Graph and Unstructured2Graph under Knowledge graph creation. - Rename Atomic Pipelines to Retrieval pipelines in the sidebar. - Split Integrations into one page per framework, keeping Integrations as a hub page. - Rename MCP to Memgraph MCP server and move the docs MCP server to its own page. - Add Coding with AI with Agent Skills, the Docs MCP server, and a section on llms.txt and Markdown pages. - Rewrite the AI ecosystem overview for the new structure and fill in missing page descriptions. Old URLs redirect to the new pages, including their .md copies. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * update: Fold the use cases into the AI ecosystem topics - Merge the Agentic GraphRAG walkthrough into the Agentic GraphRAG page as "Run it end to end". - Make AI memory and Agentic AI their own topics in AI ecosystem. - Link AI memory to the Mem0 and Cognee integrations, and Agentic AI to MemGQL Agentic Data Access in both directions. - Remove the Use cases group. Its URLs redirect to the new pages. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix: Point the Lab streams link at the Memgraph Lab page Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * update: Organize AI applications around what readers build - Rename the AI ecosystem section to AI applications. The URL stays /ai-ecosystem. - Add Start with graph + AI, with the prerequisites, the Memgraph Toolbox and knowledge graph creation. - Rename GraphRAG to Build GraphRAG. - Add Build agentic applications, with the Memgraph MCP server, AI memory and Agentic AI. - Rename Framework integrations to Integrate AI frameworks. - Move the coding assistant pages to Getting started as Use an AI coding assistant. - Move Machine learning to Advanced algorithms. - Point the Lab streams link at the Memgraph Lab page. Old URLs redirect straight to the final pages. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * update: Make GraphRAG, AI memory and Agentic AI peers in AI applications - Put AI memory and Agentic AI next to GraphRAG, matching the use cases on memgraph.com, and remove Build agentic applications. - Move the Memgraph MCP server into Tools and integrations (the former Integrate AI frameworks) and add a Memgraph Toolbox section there. - Remove the Start with graph + AI wrapper: move the prerequisites to the AI applications overview and Knowledge graph creation to the top of the section. - Use noun labels: GraphRAG, Tools and integrations. Old URLs redirect straight to the final pages. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * update: Move Memgraph Lab below Deployment Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * update: Move Machine learning back to AI applications as Graph machine learning - Put Graph machine learning between Agentic AI and Tools and integrations, at its original URL /ai-ecosystem/machine-learning, and drop the redirects to Advanced algorithms. - Link GraphRAG's use of MAGE algorithms and the gnn export module. * update: Move Help center and Release notes above Memgraph Zero * update: Describe the per-section llms.txt files on Use an AI coding assistant * fix: Accept links to the generated llms.txt files in the docs check /llms.txt and each section's llms.txt are written at build time, so a clean checkout doesn't have them. Treat them like the generated .md pages: the link passes when its section is a page or folder in pages/. * fix: Keep What's coming soon last in the sidebar The sidebar styles its last item as the "What's coming next?" button. * fix: Style the What's coming soon link as a button by its URL The sidebar button was styled with :last-child, so any item added below it broke the button. * update: Restructure GraphRAG around one name per retrieval pipeline GraphRAG follows the use-case shape (pipelines, run the example, build your own, examples). The pipelines page is Atomic GraphRAG everywhere, and its pipelines are Text2Cypher, Local graph search and Query-focused summarization on every page. Examples & demos is folded into GraphRAG and redirected. Fixes the GraphRAG skill URL and two syntax errors in the QFS summary query. * update: Make Knowledge graph creation a page for choosing how to create the graph Drops the text that repeated GraphRAG and Data modeling, adds a table from data source to tool, and names SQL2Graph and Unstructured2Graph without "Agent" to match the sidebar. * update: Rewrite the AI applications overview, AI memory and Agentic AI The overview opens with what you can build and has one-command examples. AI memory now describes what install.memgraph.com/ai-memory does since the installer hands over to Context Graph's setup. Agentic AI gets the use-case shape, numbered steps and the one-command install. * update: Fix headings, code fences and stale notes on MCP, LangChain and Graph machine learning * update: Align the AI applications overview with the reworked section pages * update: Add a how-to-choose table to Use an AI coding assistant and trim its pages The overview says when to use Agent Skills, the Docs MCP server, llms.txt, the Memgraph MCP server or AI memory, and drops Cards that repeated the list. The Docs MCP server page keeps six example prompts instead of five sections. * fix: Set --schema-info-enabled=true in the SQL2Graph start command * update: Give GraphRAG case studies their own section * update: Restore the prompt-enrichment diagram on GraphRAG and link the prerequisites * update: Describe moved AI images in alt text and place the integrations diagram under the list * update: Describe what the Agentic GraphRAG and Agentic AI examples do instead of numbering steps the script already ran * update: Fix broken code and outdated setup on the integration pages LangChain: use MemgraphLangChain and the toolkit's current five tools; the examples referenced undefined names. MCP: publish port 8000 and use the published memgraph/mcp-memgraph image. Cognee: use the community Memgraph adapter, which core Cognee no longer ships. LightRAG: install from PyPI. LlamaIndex: Markdown headings and no unsupported claims. * update: Pin the Mem0 integration to Mem0 1.x Mem0 2.0 removed graph memory from the open-source package, so mem0ai[graph] no longer installs graph support. Also fixes the Python block that contained an English sentence and drops an unsupported flag claim. * Point the Docs MCP page to the Connect AI button * Link the Connect AI dialog to the Docs MCP server page * add: Context Graph user docs under AI memory (#1847) * add: Context Graph user docs under AI memory * add: Context Graph capture privacy settings * add: Context Graph forget, uninstall and first-run setup changes * update: Context Graph plugin updates, without release versions * Update. * Remove uncessary. * update: Restructure the Context Graph docs, add a Privacy page and apply the docs rules * add: Context Graph data flow diagram on AI memory, with a dark version * update: Context Graph docs for ai-toolkit #598-#601: Agent SDK apps, config set for all keys, memory hand-back, setup's own Memgraph * update: Context Graph docs match what 0.11.0 does: fixed Explore queries, hook trust for Codex and Grok, doctor output, empty capture.connectors, entity merging, SDK plugin * update: AI memory overview is about Context Graph and how it works; ontology learning in the diagram; runnable SDK examples, credentials before setup, review fixes * update: Apply the docs writing rules to the Context Graph pages: ontology as one term, full recall and extraction key names, memory.backend default, why there are two install commands, section intros and shorter paragraphs * update: Context Graph docs for the review: knowledge graph merged into The graph, recall and extraction settings and setup flags on Configuration, Supported agents and Custom agents pages, clearer Get started steps * update: Configuration last in the AI memory sidebar and page list, Guides renamed to In this section --------- Co-authored-by: katarinasupe <supe.katarina@gmail.com> Co-authored-by: Katarina Supe <61758502+katarinasupe@users.noreply.github.com> * fix: Link the upgrade guide's rollback step to the moved HA page --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> Co-authored-by: Ante Javor <ante.javor@memgraph.io>
antejavor
added a commit
to memgraph/ai-toolkit
that referenced
this pull request
Oct 23, 2026
… the docs (#604) * memgraph-context-graph: doctor, setup and the Agent SDK match the docs Found by running every page of the AI memory docs (memgraph/documentation#1847) against 0.11.0 in a sandbox. - doctor: auto_reconcile without an LLM key no longer fails the config check; reconciliation extracts locally, the key only learns the schema. - doctor: the mcp check names only the tools the config turns on, so memory shows up only once it is the memory. - setup/update/join: reinstalling the tool keeps sessions-graph's reconciliation extra once installed, and puts the sessions-graph and resources-graph commands on PATH. Turning on resources-graph used to uninstall GLiNER2 and the sessions-graph command, and never exposed resources-graph sweep. - Agent SDK: ClaudeAgentMemory.options() turns the Context Graph Claude Code plugin off through settings. With setting_sources unset the SDK loads user settings, and the plugin's hooks recorded every prompt a second time. * context-graph: setup keeps an asked-for URL; schema learning works on anthropic 2.x Found by a second audit of the AI memory docs (memgraph/documentation#1847). - setup: a local --memgraph-url that doesn't answer (down, or refusing the login because credentials weren't set yet) is no longer swapped for setup's own container on another port; setup stops and names both ports. - sessions-graph derive: structured outputs through messages.create's output_config, which anthropic>=0.77 takes; 1.0 dropped the beta's output_format, so every candidate failed on a fresh install. derive now exits 1 when every candidate errored. - ContextGraphMemory.end_session's docstring says what ending does: embed what's left and mark the session for reconciliation.
This branch was successfully deployed
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.
Description
Add user docs for Context Graph, which gives coding agents a shared memory in Memgraph. The new section is under AI memory: an overview, then Get started, Recall and memory, Explore the graph, Privacy, Configuration, Knowledge graph, Agents, and Shared spaces (Enterprise).
Stacked on #1796. It targets that PR's branch and should be retargeted to
mainonce #1796 merges.AI memory: the overview says what Context Graph records and links to Get started instead of repeating the install steps. The Windows install now runs natively in PowerShell, so it no longer says to use WSL or Git Bash.
AI memory diagram: an animated SVG of the Context Graph data flow (capture, Memgraph, reconciliation, recall), with a dark version.
styles/globals.cssgets.light-onlyand.dark-onlyclasses to show one image per theme. Readers with reduced motion see it static.Enabling Memgraph Enterprise: lists Context Graph shared spaces.
Privacy: where recorded sessions go (only the configured Memgraph, except that schema learning with an LLM API key sends full session text to that provider), and a "Keep text out of the graph" section that covers the capture privacy settings from memgraph-context-graph: redact secrets, strip <private> and honour capture exclusions before connectors store events ai-toolkit#585:
<private>blocks, secret redaction and[redact]patterns,capture.max_chars,capture.exclude_toolsandcapture.exclude_projects.Privacy: Forget sessions (context-graph: forget sessions for good, each component deleting its own share ai-toolkit#588).
Configuration: the reworked uninstall with
--graph,--config,--tooland--all(memgraph-context-graph: uninstall asks whether the graph, config and tool go too; --all removes everything ai-toolkit#590).Configuration: how
updaterefreshes the packages, plugins, hook files and schema, and how to update the Claude Code and Codex plugins by hand.Get started: on a first run, setup ticks only Claude Code and Codex; the
doctoroutput shows theplugin:check (context-graph: setup wires the agents a new user has, and the install matrix proves a real session lands ai-toolkit#589).Not verified yet:
uv tool install … --with-executables-from sessions-graphcommand on the Knowledge graph page, which installs the extraction packages, hasn't been run.started_at,text,entity_type). They haven't been run against a captured graph.joinoutput on Shared spaces comes from the command's print statements, not from a real run.Product PR
None. These pages document memgraph/ai-toolkit's
context-graph.Checklist
featureorbugfixlabel added