@@ -64,6 +64,12 @@ Every extractor returns:
 
 `validate.py` enforces this schema before `build()` consumes it.
 
+## Edge direction in graph.json
+
+`graph.json` is written `"directed": false`, but every link is directed: **arc order is the direction** (`source` → `target`, #563). Older canonicalized files carry it in `_src`/`_tgt` markers instead, and those win where present (#2309).
+
+A plain `json_graph.node_link_graph()` load loses this: an undirected graph re-orders each edge's endpoints by node-list position. Load through `paths.load_node_link_graph()` (or apply `paths.restore_arc_direction()` to the raw dict first), which stamps `_src`/`_tgt` on every edge. Then read direction as `data.get("_src", u)` / `data.get("_tgt", v)`, never from `u, v`. Anything that writes graph.json back out pops the markers into arc order first, as `export.to_json()` does.
+
 ## Confidence labels
 
 | Label | Meaning |
@@ -93,10 +99,22 @@ See `SECURITY.md` for the full threat model.
 
 ## Testing
 
+`analyze.suggest_questions()` interleaves candidates across question types before
+applying its result limit (seven by default). This keeps ambiguous relationships
+from hiding bridge, inferred-relationship, isolation, and low-cohesion signals.
+The existing candidate order within each type is preserved; if the limit is smaller
+than the number of available types, their generation order determines priority.
+
 One test file per module under `tests/`. Run with:
 
 ```bash
 pytest tests/ -q
 ```
 
 The test suite is designed to avoid network access and uncontrolled filesystem effects; most tests are isolated with temporary directories and environment fixtures.
+
+R extraction preserves symbol-only bindings such as `%||%` as separate nodes.
+When normal ID normalization removes an entire binding name, its UTF-8 bytes
+provide a deterministic operator suffix within the same lexical owner. Its double
+underscore separators cannot collide with an ordinary normalized binding ID. Local
+call resolution stops if malformed scope metadata cycles.
@@ -30,6 +30,7 @@ Highlights:
 | Memory | LOCOMO (300) | QA accuracy | 45.3% | supermemory 49.7% (11x ingest cost), bm25 31.3%, mem0 27.3% |
 | Memory | LOCOMO (300) | recall@10 | 0.497 | bm25 0.362, mem0 0.048 |
 | Memory | LongMemEval-S (50) | QA accuracy | 76% | dense RAG 76%, hybrid 74%, mem0 70% |
+| Code | ERPNext cross-tool (6) | key-fact coverage | 82.0% | grep/read baseline 70.8% |
 | Cost | LOCOMO ingest | USD | ~$1.40 | supermemory $15.67, mem0 $3.48 |
 | Cost | graph build | LLM credits | $0 | n/a |
 
@@ -148,6 +149,21 @@ graphify pays for itself in accuracy against searching raw files, and avoids the
 context-stuffing anti-pattern of packing the whole repo into every turn (which
 costs roughly 20x the tokens for lower coverage).
 
+### INFERRED cross-file edge precision (FastAPI demo corpus)
+
+The README hero graph is the upstream [FastAPI](https://github.com/fastapi/fastapi)
+library extracted AST-only (no semantic pass). To check the `INFERRED` confidence
+tag on cross-file resolver edges, we manually audited every such edge
+(`calls`, `indirect_call`, `imports`, `imports_from`, `inherits`) whose source and
+target both live under `fastapi/` (production package sources, excluding the test
+tree). This is a dated manual spot-check, not a continuously-measured metric: on a
+fresh AST-only extract (2026-10-05) all 12 such edges were correct. Reproduce by
+running `graphify .` against a FastAPI checkout and inspecting the INFERRED
+cross-file edges whose endpoints both sit under `fastapi/`. Homonymous helpers in
+`tests/` produce noisier INFERRED `calls` edges; the tag is most meaningful on
+production cross-file resolution, where the resolver prefers no edge over a wrong
+one.
+
 ## Results: temporal (15 years of ERPNext)
 
 689 weekly AST checkpoints, 2011 to 2026, built deterministically with no LLM.
 
@@ -42,8 +42,9 @@ git remote add upstream https://github.com/Graphify-Labs/graphify.git
 git fetch upstream
 git checkout -b my-feature-branch upstream/v8  # Always branch off upstream v8, never commit directly to v8
 
-# Set up the environment
-uv sync
+# Set up the environment (--all-extras installs the optional grammars and SDKs,
+# matching CI; without it the tests that need them are skipped)
+uv sync --all-extras
 uv run pre-commit install
 
 # Run the test suite to confirm your setup
 
@@ -20,14 +20,15 @@
   <a href="LICENSE"><img src="https://img.shields.io/badge/License-Apache%202.0-blue" alt="License: Apache-2.0"/></a>
   <a href="https://pepy.tech/project/graphifyy"><img src="https://img.shields.io/pepy/dt/graphifyy?color=blue&label=downloads" alt="Downloads"/></a>
   <a href="https://docs.graphify.com"><img src="https://img.shields.io/badge/Docs-docs.graphify.com-0b7285?style=flat&logo=readthedocs&logoColor=white" alt="Docs"/></a>
+  <a href="https://smithery.ai/servers/graphify/graphify"><img src="https://img.shields.io/badge/Smithery-MCP%20server-5c4ee5?style=flat&logo=data:image/svg+xml;base64,PHN2ZyB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciIHZpZXdCb3g9IjAgMCAyNCAyNCIgZmlsbD0iI2ZmZiI+PHBhdGggZD0iTTEyIDJMMiA3djEwbDEwIDUgMTAtNVY3eiIvPjwvc3ZnPg==" alt="Smithery"/></a>
   <a href="https://discord.gg/XDnKVpzdXB"><img src="https://img.shields.io/badge/Discord-Join-5865F2?style=flat&logo=discord&logoColor=white" alt="Discord"/></a>
   <a href="https://www.youtube.com/@graphifylabs"><img src="https://img.shields.io/badge/YouTube-Graphify%20Labs-FF0000?style=flat&logo=youtube&logoColor=white" alt="YouTube"/></a>
   <a href="https://www.linkedin.com/company/graphify-labs"><img src="https://img.shields.io/badge/LinkedIn-Graphify%20Labs-0077B5?logo=linkedin" alt="LinkedIn"/></a>
   <a href="https://www.ycombinator.com/companies/graphify-labs"><img src="https://img.shields.io/badge/Y%20Combinator-S26-F0652F?style=flat&logo=ycombinator&logoColor=white" alt="YC S26"/></a>
 </p>
 
 <p align="center">
-  <b>Early access to the graphify platform is open before the public v1 launch: <a href="https://app.graphify.com/login">app.graphify.com</a></b>
+  <b>Try the graphify platform free for 14 days: <a href="https://app.graphify.com/login">app.graphify.com</a></b>
 </p>
 
 Type `/graphify` in your AI coding assistant and it maps your entire project (code, docs, PDFs, images, videos) into a **knowledge graph** you can **query instead of grepping** through files.
@@ -67,6 +68,10 @@ graphify-out/
 └── graph.json       the full graph — query it anytime without re-reading your files
 ```
 
+The persisted graph includes `graph.schema_version` so integrations can detect
+incompatible format changes, plus `graph.graphify_version` identifying the
+Graphify release that produced it.
+
 **Works in** Claude Code, Cursor, Codex, Gemini CLI, GitHub Copilot, and 15+ more — [pick your platform](#install).
 
 ---
@@ -125,6 +130,7 @@ What you get out of the box:
 | LOCOMO (n=300) | recall@10 | **0.497** | mem0 0.048, supermemory 0.149 |
 | LOCOMO (n=300) | QA accuracy | 45.3% | supermemory 49.7%, mem0 27.3% |
 | LongMemEval-S (n=50) | QA accuracy | **76%** | tied with dense RAG |
+| ERPNext cross-tool (n=6) | key-fact coverage | **82.0%** | grep/read baseline 70.8% |
 | Graph build | LLM credits | **0** | per-token for most systems |
 
 Every system ran on the same harness with the same model and budgets, scored by a judge blind-validated against a second judge (90.6% agreement, Cohen's kappa 0.81). Full per-system tables, the code-intelligence result, and reproduction commands: **[BENCHMARKS.md](./BENCHMARKS.md)**.
@@ -344,7 +350,7 @@ To remove graphify from all platforms at once: `graphify uninstall` (add `--purg
 
 | Type | Extensions |
 |------|-----------|
-| Code (37 tree-sitter grammars) | `.py .ts .mts .cts .js .jsx .tsx .mjs .go .rs .java .c .cpp .cc .cxx .h .hpp .cu .cuh .metal .rb .cs .kt .kts .scala .php .swift .lua .luau .toc .zig .ps1 .psm1 .psd1 .ex .exs .m .mm .ml .mli .jl .vue .svelte .astro .groovy .gradle .dart .v .sv .svh .sql .f .f90 .f95 .f03 .f08 .pas .pp .dpr .dpk .lpr .inc .dfm .lfm .lpk .sh .bash .json .dm .dme .dmi .dmm .dmf .sln .slnx .csproj .fsproj .vbproj .xaml .razor .cshtml` (`.dm`/`.dme` requires `uv tool install graphifyy[dm]`, `.ml`/`.mli` requires `uv tool install graphifyy[ocaml]`; `.mts`/`.cts` reuse the TypeScript grammar, `.cc`/`.cxx` and CUDA `.cu`/`.cuh` and Metal `.metal` reuse the C++ grammar) |
+| Code (37 tree-sitter grammars) | `.py .ts .mts .cts .js .jsx .tsx .mjs .go .rs .java .c .cpp .cc .cxx .h .hpp .cu .cuh .metal .rb .cs .kt .kts .scala .php .swift .lua .luau .toc .zig .ps1 .psm1 .psd1 .ex .exs .m .mm .ml .mli .jl .vue .svelte .astro .groovy .gradle .dart .v .sv .svh .vh .sql .f .f90 .f95 .f03 .f08 .pas .pp .dpr .dpk .lpr .inc .dfm .lfm .lpk .sh .bash .json .dm .dme .dmi .dmm .dmf .sln .slnx .csproj .fsproj .vbproj .xaml .razor .cshtml` (`.dm`/`.dme` requires `uv tool install graphifyy[dm]`, `.ml`/`.mli` requires `uv tool install graphifyy[ocaml]`; `.mts`/`.cts` reuse the TypeScript grammar, `.cc`/`.cxx` and CUDA `.cu`/`.cuh` and Metal `.metal` reuse the C++ grammar) |
 | Salesforce Apex | `.cls .trigger` (regex-based; classes, interfaces, enums, methods, triggers, SOQL/DML edges) |
 | Terraform / HCL | `.tf .tfvars .hcl` (requires `uv tool install graphifyy[terraform]`) |
 | OCaml | `.ml .mli` (requires `uv tool install graphifyy[ocaml]`) |
@@ -512,6 +518,19 @@ python -m graphify.serve graphify-out/graph.json --transport http --host 0.0.0.0
 
 The MCP server gives your assistant structured access: `query_graph`, `get_node`, `get_neighbors`, `shortest_path`, `list_prs`, `get_pr_impact`, `triage_prs`.
 
+### One-click install (VS Code · Smithery)
+
+Prefer not to edit MCP config by hand? Register the graphify MCP server in one step. Run `/graphify` first so a `graphify-out/graph.json` exists in the workspace; the server reads it on launch.
+
+[![Install in VS Code](https://img.shields.io/badge/VS_Code-Install_graphify_MCP-0098FF?style=flat-square&logo=visualstudiocode&logoColor=white)](https://insiders.vscode.dev/redirect/mcp/install?name=graphify&config=%7B%22command%22%3A%22uvx%22%2C%22args%22%3A%5B%22--from%22%2C%22graphifyy%5Bmcp%5D%22%2C%22graphify-mcp%22%5D%7D)
+
+The button adds a `graphify` server that launches with `uvx --from "graphifyy[mcp]" graphify-mcp` (no API key, no prior install). Or install via [Smithery](https://smithery.ai/servers/graphify/graphify) for Claude, Cursor, VS Code, and other clients:
+
+```bash
+# pick your client (claude, cursor, vscode, windsurf, ...)
+npx -y @smithery/cli install graphify/graphify --client claude
+```
+
 ### Shared HTTP server
 
 `--transport stdio` (the default) spawns one local server per developer. `--transport http` serves the same tools over the MCP Streamable HTTP transport, so a single shared process can serve the graph for the whole team — clients point their IDE MCP config at `http://<host>:8080/mcp` instead of running graphify locally.
@@ -579,6 +598,7 @@ These are only needed for **headless / CI extraction** (`graphify extract`). Whe
 | `GRAPHIFY_QUERY_LOG` | Enable the query log and write it to this path instead of the default | optional — off unless this or `_ENABLE` is set |
 | `GRAPHIFY_QUERY_LOG_DISABLE` | Set to `1` to force the query log off (wins over the enable vars) | optional |
 | `GRAPHIFY_QUERY_LOG_RESPONSES` | When the log is enabled, also record full subgraph responses (off by default) | optional |
+| `GRAPHIFY_NO_AUTO_REFRESH` | Set to `1` to stop the CLI from refreshing installed skills that are older than the package after an upgrade | optional — refresh is on by default |
 | `GRAPHIFY_MAX_GRAPH_BYTES` | Override the 512 MiB graph.json size cap — e.g. `700MB`, `2GB`, or plain bytes | optional — useful for very large corpora |
 | `GRAPHIFY_MAX_CONTEXTS` | Maximum number of non-default project graphs retained by one multi-project MCP server | optional — default: `8`; invalid values use `8`, and values below `1` use `1` |
 | `GRAPHIFY_LLM_TEMPERATURE` | Override LLM temperature for semantic extraction — e.g. `0.7`, or `none` to omit | optional — auto-omitted for o1/o3/o4/gpt-5 reasoning models |
@@ -674,10 +694,10 @@ ANTHROPIC_API_KEY=sk-... graphify extract ./docs --backend claude
 ```
 
 **Skill version mismatch warning in your IDE**
-Your installed graphify version is different from the skill file. Update:
+Your installed graphify version is different from the skill file. After an upgrade, the first `graphify` command refreshes every installed skill that is older than the package, on all platforms at once, and keeps a locally edited `SKILL.md` as `SKILL.md.bak`. Set `GRAPHIFY_NO_AUTO_REFRESH=1` to turn this off. A skill it can't refresh safely keeps the warning: one newer than the package, or a directory that two installers share (for example Copilot and `graphify vscode install`). Reinstall that platform by hand:
 ```bash
 uv tool upgrade graphifyy
-graphify install  # overwrites the skill file
+graphify install --platform <name>  # overwrites the skill file
 ```
 
 **Claude Code prompt cache invalidated after every `graphify extract`**
@@ -845,6 +865,12 @@ graphify label ./my-project                                    # (re)name commun
 graphify label ./my-project --backend=openai --model gpt-4o   # force a specific backend and model
 ```
 
+`--no-dedup` also skips coalescing distinct non-AST nodes solely because they
+share a file and label. The Python equivalents are `build(chunks, dedup=False)`,
+`build_merge(chunks, graph_path, dedup=False)`, and
+`build_from_json(extraction, dedup=False)`. AST/semantic twins still reconcile to
+the canonical AST node, and document-file twin reconciliation remains enabled.
+
 > **Community names:** inside an agent (Claude Code, Gemini CLI) the agent names communities itself. When you run the bare CLI, `cluster-only` auto-names them with the configured backend (built-in or custom OpenAI-compatible provider) — pass `--no-label` to keep `Community N`, or run `graphify label` to (re)generate names on demand.
 
 ---
@@ -877,6 +903,22 @@ New here? Say hi on [Discord](https://discord.gg/XDnKVpzdXB) or in [GitHub Discu
 
 ---
 
+## Contributors
+
+<a href="https://github.com/Graphify-Labs/graphify/graphs/contributors">
+  <img src="https://contrib.rocks/image?repo=Graphify-Labs/graphify" alt="graphify contributors" />
+</a>
+
+Made with [contrib.rocks](https://contrib.rocks).
+
+---
+
+## Translations
+
+The README is available in 32 languages. Use the language switcher at the top of this file to read it in yours, or browse [`docs/translations/`](docs/translations/). To improve a translation or add a new one, open a pull request against the matching file there.
+
+---
+
 ## Community and links
 
 <p align="center">
 
@@ -48,9 +48,9 @@ EXTRACTED edges always have confidence 1.0. INFERRED edges use a discrete rubric
 - **0.65** — weak (naming similarity only)
 - **0.55** — speculative
 
-That rubric describes edges Claude inferred. The per-language member-call resolvers are a separate, deterministic source of INFERRED edges: they read the receiver's declared type out of the AST and bind at a fixed **0.8**, reserving EXTRACTED (1.0) for a receiver whose type is named in the source at the call site.
+That rubric describes edges Claude inferred. The per-language member-call resolvers are a separate, deterministic source of INFERRED edges: they read the receiver's declared type out of the AST and bind at a fixed **0.85** (a step of the same rubric), reserving EXTRACTED (1.0) for a receiver whose type is named in the source at the call site.
 
-**PHP member calls refuse rather than guess.** `$this->prop->method()`, `$obj?->method()`, a typed parameter and a `$var = new T()` local all bind to the receiver's declared type as INFERRED 0.8; `(new Service())->method()` is EXTRACTED 1.0, but only when the namespace written at the call site corroborates the class that was found. When the type is not provably one concrete in-corpus class, no edge is emitted at all — untyped, union- and intersection-typed receivers, a short name that matches two classes, a method the receiver's class does not declare, chained and array-element receivers, a local rebound or shadowed anywhere in the method, anonymous classes, and `self`/`static`/`parent`. A receiver typed by an `interface`, `enum` or `trait` binds to that declaration's own method only; implementations are never guessed — `MailNotifier implements Notifier` gets nothing from a `Notifier`-typed receiver. A Laravel corpus has many identically named `search()`/`log()`/`handle()` methods, so an absent edge is worth more than a guessed one.
+**PHP member calls refuse rather than guess.** `$this->prop->method()`, `$obj?->method()`, a typed parameter and a `$var = new T()` local all bind to the receiver's declared type as INFERRED 0.85; `(new Service())->method()` is EXTRACTED 1.0, but only when the namespace written at the call site corroborates the class that was found. When the type is not provably one concrete in-corpus class, no edge is emitted at all — untyped, union- and intersection-typed receivers, a short name that matches two classes, a method the receiver's class does not declare, chained and array-element receivers, a local rebound or shadowed anywhere in the method, anonymous classes, and `self`/`static`/`parent`. A receiver typed by an `interface`, `enum` or `trait` binds to that declaration's own method only; implementations are never guessed — `MailNotifier implements Notifier` gets nothing from a `Notifier`-typed receiver. A Laravel corpus has many identically named `search()`/`log()`/`handle()` methods, so an absent edge is worth more than a guessed one.
 
 A PHP 8.1 first-class callable — `$obj->method(...)`, including the nullsafe and `$this` forms — resolves by exactly those rules but is emitted as **`indirect_call`**, not `calls`: the syntax creates a `Closure` and names the method without invoking it, the same shape the resolver already labels `indirect_call` for a callback passed by name. `calls` therefore keeps meaning "control flow transfers here". The spread form `$obj->method(...$args)` is a real invocation and stays `calls`.