d365fo-agent-developer 0.6.0__tar.gz → 0.6.2__tar.gz

{d365fo_agent_developer-0.6.0/src/d365fo_agent_developer.egg-info → d365fo_agent_developer-0.6.2}/PKG-INFO

@@ -1,12 +1,12 @@
 Metadata-Version: 2.4
 Name: d365fo-agent-developer
-Version: 0.6.0
+Version: 0.6.2
 Summary: Local D365 F&O knowledge toolkit: corpus indexing (SQLite FTS5), deterministic generation, and an MCP server that grounds coding agents (Claude Code, Codex) in real AOT facts.
 Author-email: David Bru <dbru@fiveforty.fr>
 License: MIT
-Project-URL: Homepage, https://github.com/dbru540/d365fo-agent
-Project-URL: Repository, https://github.com/dbru540/d365fo-agent
-Project-URL: Issues, https://github.com/dbru540/d365fo-agent/issues
+Project-URL: Homepage, https://github.com/dbru540/d365fo-agent-developer
+Project-URL: Repository, https://github.com/dbru540/d365fo-agent-developer
+Project-URL: Issues, https://github.com/dbru540/d365fo-agent-developer/issues
 Keywords: dynamics-365,d365fo,x++,aot,mcp,model-context-protocol,claude,claude-code,codex,ai,code-generation,rag
 Classifier: Development Status :: 4 - Beta
 Classifier: Intended Audience :: Developers
@@ -29,7 +29,7 @@ Requires-Dist: ruff; extra == "dev"
 Requires-Dist: build; extra == "dev"
 Dynamic: license-file
 
-# d365fo-agent — D365 F&O X++ knowledge for Claude Code & Codex
+# d365fo-agent-developer — D365 F&O X++ knowledge for Claude Code & Codex
 
 A local **MCP server** that gives an AI coding agent (Claude Code, Codex, any MCP host) a grounded
 **knowledge base of Dynamics 365 Finance & Operations X++** — so you can develop for D365 quickly
@@ -51,9 +51,14 @@ compiler.
 ```bash
 pip install d365fo-agent-developer
 # or, isolated:  pipx install d365fo-agent-developer
+# or zero-install, straight from PyPI at launch time:  uvx d365fo-agent-developer
 ```
 
-This installs two commands: `d365fo-mcp` (the server) and `d365fo-agent` (the CLI).
+With `uvx` you can even skip the install entirely — point your MCP client at
+`uvx` with args `["d365fo-agent-developer"]` (e.g. in Claude Desktop's
+`claude_desktop_config.json`) and the latest published server runs on demand.
+
+This installs two commands: `d365fo-agent-developer` (the MCP server; `d365fo-mcp` is a short alias) and `d365fo-agent` (the CLI).
 
 ## Get the knowledge base (once)
 
@@ -84,18 +89,18 @@ d365fo-agent build-index \
 ```json
 {
   "mcpServers": {
-    "d365fo": { "command": "d365fo-mcp", "args": [] }
+    "d365fo-agent-developer": { "command": "d365fo-agent-developer", "args": [] }
   }
 }
 ```
 
-…or one command: `claude mcp add d365fo d365fo-mcp`
+…or one command: `claude mcp add d365fo-agent-developer d365fo-agent-developer`
 
 **Codex** — add to `~/.codex/config.toml`:
 
 ```toml
-[mcp_servers.d365fo]
-command = "d365fo-mcp"
+[mcp_servers.d365fo-agent-developer]
+command = "d365fo-agent-developer"
 args = []
 ```
 
@@ -109,8 +114,8 @@ Point the server at your D365 source repo so your **custom** classes/tables/EDTs
 are indexed too, and so the rich tools can read real signatures and clone real examples:
 
 ```toml
-[mcp_servers.d365fo]
-command = "d365fo-mcp"
+[mcp_servers.d365fo-agent-developer]
+command = "d365fo-agent-developer"
 args = ["--repo-root", "C:/path/to/your/D365Repo",
         "--rules", "C:/path/to/your/rules.json",
         "--packages-root", "C:/AOSService/PackagesLocalDirectory"]
@@ -127,7 +132,8 @@ args = ["--repo-root", "C:/path/to/your/D365Repo",
 `element_exists`, `find_element`, `search_corpus`, `get_signature`, `get_extension_chain`,
 `get_security_links`, `get_entity_exposure`, `find_similar_examples`, `scaffold_object`,
 `find_references`, `find_reverse_references`, `analyze_spec`, `generate_from_spec`, `validate_xml`,
-`lint_artifact`, `derive_entity`, `wire_security`, `compile_model`, `get_methodology`, `index_stats`.
+`lint_artifact`, `derive_entity`, `wire_security`, `compile_model`, `compile_generated`,
+`get_sql_model`, `explore_functional_unit`, `find_relations`, `get_methodology`, `index_stats`.
 
 See [docs/mcp-server.md](docs/mcp-server.md) for the verify-driven workflow and
 [docs/x++-methodology.md](docs/x++-methodology.md) for the behavioural contract.

{d365fo_agent_developer-0.6.0 → d365fo_agent_developer-0.6.2}/README.md

@@ -1,140 +1,146 @@
-# d365fo-agent — D365 F&O X++ knowledge for Claude Code & Codex
-
-A local **MCP server** that gives an AI coding agent (Claude Code, Codex, any MCP host) a grounded
-**knowledge base of Dynamics 365 Finance & Operations X++** — so you can develop for D365 quickly
-**without re-feeding a whole repo for analysis every time**.
-
-It grounds the agent in real AOT facts: it verifies that a class/table/EDT/enum/entity actually
-exists (anti-hallucination), searches the corpus, walks extension/security relationships, serves an
-X++ engineering methodology, validates generated XML against learned per-type structure, scaffolds
-or deterministically generates artifacts, and (on a Windows D365 host) compiles with the real X++
-compiler.
-
-- **Pure Python standard library** — zero runtime dependencies, runs anywhere Python 3.11+ runs.
-- **Vendor-neutral** — Claude Code, Codex, Gemini CLI, or any MCP-speaking client over stdio.
-- The standard D365 corpus is the same for everyone, so it is indexed **once** and used as a
-  portable knowledge base; your own custom code is **optional**.
-
-## Install
-
-```bash
-pip install d365fo-agent-developer
-# or, isolated:  pipx install d365fo-agent-developer
-```
-
-This installs two commands: `d365fo-mcp` (the server) and `d365fo-agent` (the CLI).
-
-## Get the knowledge base (once)
-
-Pick one. Both produce a local index at `~/.d365fo-agent/d365fo.db` that the server uses by default.
-
-**A. Download the prebuilt standard-D365 index** (fastest, no D365 install needed):
-
-```bash
-d365fo-agent fetch-knowledge          # downloads + caches the standard knowledge index
-```
-
-**B. Build it from your own D365 dev box** (no download; uses metadata you already have):
-
-```bash
-d365fo-agent build-index \
-  --db ~/.d365fo-agent/d365fo.db \
-  --packages-root "C:/AOSService/PackagesLocalDirectory" \
-  --rebuild
-```
-
-> The methodology, default lint rules, and a default learned type-profile ship **inside** the
-> package, so validation and guidance work out of the box even before the index is built.
-
-## Wire it into your agent
-
-**Claude Code** — add to `.mcp.json` (project) or `~/.claude.json` (global):
-
-```json
-{
-  "mcpServers": {
-    "d365fo": { "command": "d365fo-mcp", "args": [] }
-  }
-}
-```
-
-…or one command: `claude mcp add d365fo d365fo-mcp`
-
-**Codex** — add to `~/.codex/config.toml`:
-
-```toml
-[mcp_servers.d365fo]
-command = "d365fo-mcp"
-args = []
-```
-
-With no `--db`, the server uses the cached knowledge index automatically. That's it — ask the agent
-to build something for D365 and it will verify elements, follow the methodology, and validate its
-output instead of guessing.
-
-## Add your custom code (optional)
-
-Point the server at your D365 source repo so your **custom** classes/tables/EDTs/enums/extensions
-are indexed too, and so the rich tools can read real signatures and clone real examples:
-
-```toml
-[mcp_servers.d365fo]
-command = "d365fo-mcp"
-args = ["--repo-root", "C:/path/to/your/D365Repo",
-        "--rules", "C:/path/to/your/rules.json",
-        "--packages-root", "C:/AOSService/PackagesLocalDirectory"]
-```
-
-| Capability | Knowledge index only | + a PackagesLocalDirectory / repo |
-|---|---|---|
-| Verify an element exists, search, relations, methodology, validation, scaffolding by template | ✅ | ✅ |
-| Read a real signature, clone a real example (`get_signature`, `find_similar_examples`, `scaffold_object`) | needs source files | ✅ |
-| Compile with the real X++ compiler (`compile_model`) | — | ✅ (Windows D365 host) |
-
-## What the agent gets (MCP tools)
-
-`element_exists`, `find_element`, `search_corpus`, `get_signature`, `get_extension_chain`,
-`get_security_links`, `get_entity_exposure`, `find_similar_examples`, `scaffold_object`,
-`find_references`, `find_reverse_references`, `analyze_spec`, `generate_from_spec`, `validate_xml`,
-`lint_artifact`, `derive_entity`, `wire_security`, `compile_model`, `get_methodology`, `index_stats`.
-
-See [docs/mcp-server.md](docs/mcp-server.md) for the verify-driven workflow and
-[docs/x++-methodology.md](docs/x++-methodology.md) for the behavioural contract.
-
-## Maintainer: publish the knowledge index
-
-The wheel stays tiny; the ~100 MB standard index is distributed as a downloadable asset.
-
-```bash
-# 1. Build a STANDARD-only index from a PackagesLocalDirectory (no custom repo)
-d365fo-agent build-index --db d365fo-standard.db --packages-root <PLD> --rebuild
-# 2. (optional) learn type profiles to ship as the default
-d365fo-agent build-type-profiles --db d365fo-standard.db --packages-root <PLD> \
-  --out src/d365fo_agent/data/aot-type-profiles.json
-# 3. Compress and attach to a GitHub release
-python -c "import gzip,shutil; shutil.copyfileobj(open('d365fo-standard.db','rb'), gzip.open('d365fo-standard.db.gz','wb'))"
-# 4. Point users at it: set DEFAULT_KNOWLEDGE_URL in knowledge_fetch.py (or pass --url)
-```
-
-> **Note:** the index holds factual AOT metadata (element names, types, packages, labels,
-> relations) — not Microsoft source. Confirm your redistribution position before publishing a
-> prebuilt standard index; option **B** above lets each user build their own with zero redistribution.
-
-To publish the package itself: `python -m build` then `python -m twine upload dist/*` (PyPI account
-required).
-
-## Develop / contribute
-
-```bash
-pip install -e ".[dev]"
-PYTHONPATH=src python -m unittest discover -s tests   # full test suite
-ruff check src/d365fo_agent tests
-```
-
-Docs: [Architecture](docs/architecture.md) · [MCP Server](docs/mcp-server.md) ·
-[X++ Methodology](docs/x++-methodology.md) · [Specification Contract](docs/specification-contract.md) ·
-[Metadata Schema](docs/metadata-schema.md) · [Tool Catalog](docs/tool-catalog.md)
-
-## License
-
-MIT — see [LICENSE](LICENSE).
+# d365fo-agent-developer — D365 F&O X++ knowledge for Claude Code & Codex
+
+A local **MCP server** that gives an AI coding agent (Claude Code, Codex, any MCP host) a grounded
+**knowledge base of Dynamics 365 Finance & Operations X++** — so you can develop for D365 quickly
+**without re-feeding a whole repo for analysis every time**.
+
+It grounds the agent in real AOT facts: it verifies that a class/table/EDT/enum/entity actually
+exists (anti-hallucination), searches the corpus, walks extension/security relationships, serves an
+X++ engineering methodology, validates generated XML against learned per-type structure, scaffolds
+or deterministically generates artifacts, and (on a Windows D365 host) compiles with the real X++
+compiler.
+
+- **Pure Python standard library** — zero runtime dependencies, runs anywhere Python 3.11+ runs.
+- **Vendor-neutral** — Claude Code, Codex, Gemini CLI, or any MCP-speaking client over stdio.
+- The standard D365 corpus is the same for everyone, so it is indexed **once** and used as a
+  portable knowledge base; your own custom code is **optional**.
+
+## Install
+
+```bash
+pip install d365fo-agent-developer
+# or, isolated:  pipx install d365fo-agent-developer
+# or zero-install, straight from PyPI at launch time:  uvx d365fo-agent-developer
+```
+
+With `uvx` you can even skip the install entirely — point your MCP client at
+`uvx` with args `["d365fo-agent-developer"]` (e.g. in Claude Desktop's
+`claude_desktop_config.json`) and the latest published server runs on demand.
+
+This installs two commands: `d365fo-agent-developer` (the MCP server; `d365fo-mcp` is a short alias) and `d365fo-agent` (the CLI).
+
+## Get the knowledge base (once)
+
+Pick one. Both produce a local index at `~/.d365fo-agent/d365fo.db` that the server uses by default.
+
+**A. Download the prebuilt standard-D365 index** (fastest, no D365 install needed):
+
+```bash
+d365fo-agent fetch-knowledge          # downloads + caches the standard knowledge index
+```
+
+**B. Build it from your own D365 dev box** (no download; uses metadata you already have):
+
+```bash
+d365fo-agent build-index \
+  --db ~/.d365fo-agent/d365fo.db \
+  --packages-root "C:/AOSService/PackagesLocalDirectory" \
+  --rebuild
+```
+
+> The methodology, default lint rules, and a default learned type-profile ship **inside** the
+> package, so validation and guidance work out of the box even before the index is built.
+
+## Wire it into your agent
+
+**Claude Code** — add to `.mcp.json` (project) or `~/.claude.json` (global):
+
+```json
+{
+  "mcpServers": {
+    "d365fo-agent-developer": { "command": "d365fo-agent-developer", "args": [] }
+  }
+}
+```
+
+…or one command: `claude mcp add d365fo-agent-developer d365fo-agent-developer`
+
+**Codex** — add to `~/.codex/config.toml`:
+
+```toml
+[mcp_servers.d365fo-agent-developer]
+command = "d365fo-agent-developer"
+args = []
+```
+
+With no `--db`, the server uses the cached knowledge index automatically. That's it — ask the agent
+to build something for D365 and it will verify elements, follow the methodology, and validate its
+output instead of guessing.
+
+## Add your custom code (optional)
+
+Point the server at your D365 source repo so your **custom** classes/tables/EDTs/enums/extensions
+are indexed too, and so the rich tools can read real signatures and clone real examples:
+
+```toml
+[mcp_servers.d365fo-agent-developer]
+command = "d365fo-agent-developer"
+args = ["--repo-root", "C:/path/to/your/D365Repo",
+        "--rules", "C:/path/to/your/rules.json",
+        "--packages-root", "C:/AOSService/PackagesLocalDirectory"]
+```
+
+| Capability | Knowledge index only | + a PackagesLocalDirectory / repo |
+|---|---|---|
+| Verify an element exists, search, relations, methodology, validation, scaffolding by template | ✅ | ✅ |
+| Read a real signature, clone a real example (`get_signature`, `find_similar_examples`, `scaffold_object`) | needs source files | ✅ |
+| Compile with the real X++ compiler (`compile_model`) | — | ✅ (Windows D365 host) |
+
+## What the agent gets (MCP tools)
+
+`element_exists`, `find_element`, `search_corpus`, `get_signature`, `get_extension_chain`,
+`get_security_links`, `get_entity_exposure`, `find_similar_examples`, `scaffold_object`,
+`find_references`, `find_reverse_references`, `analyze_spec`, `generate_from_spec`, `validate_xml`,
+`lint_artifact`, `derive_entity`, `wire_security`, `compile_model`, `compile_generated`,
+`get_sql_model`, `explore_functional_unit`, `find_relations`, `get_methodology`, `index_stats`.
+
+See [docs/mcp-server.md](docs/mcp-server.md) for the verify-driven workflow and
+[docs/x++-methodology.md](docs/x++-methodology.md) for the behavioural contract.
+
+## Maintainer: publish the knowledge index
+
+The wheel stays tiny; the ~100 MB standard index is distributed as a downloadable asset.
+
+```bash
+# 1. Build a STANDARD-only index from a PackagesLocalDirectory (no custom repo)
+d365fo-agent build-index --db d365fo-standard.db --packages-root <PLD> --rebuild
+# 2. (optional) learn type profiles to ship as the default
+d365fo-agent build-type-profiles --db d365fo-standard.db --packages-root <PLD> \
+  --out src/d365fo_agent/data/aot-type-profiles.json
+# 3. Compress and attach to a GitHub release
+python -c "import gzip,shutil; shutil.copyfileobj(open('d365fo-standard.db','rb'), gzip.open('d365fo-standard.db.gz','wb'))"
+# 4. Point users at it: set DEFAULT_KNOWLEDGE_URL in knowledge_fetch.py (or pass --url)
+```
+
+> **Note:** the index holds factual AOT metadata (element names, types, packages, labels,
+> relations) — not Microsoft source. Confirm your redistribution position before publishing a
+> prebuilt standard index; option **B** above lets each user build their own with zero redistribution.
+
+To publish the package itself: `python -m build` then `python -m twine upload dist/*` (PyPI account
+required).
+
+## Develop / contribute
+
+```bash
+pip install -e ".[dev]"
+PYTHONPATH=src python -m unittest discover -s tests   # full test suite
+ruff check src/d365fo_agent tests
+```
+
+Docs: [Architecture](docs/architecture.md) · [MCP Server](docs/mcp-server.md) ·
+[X++ Methodology](docs/x++-methodology.md) · [Specification Contract](docs/specification-contract.md) ·
+[Metadata Schema](docs/metadata-schema.md) · [Tool Catalog](docs/tool-catalog.md)
+
+## License
+
+MIT — see [LICENSE](LICENSE).

{d365fo_agent_developer-0.6.0 → d365fo_agent_developer-0.6.2}/pyproject.toml

@@ -1,57 +1,59 @@
-[build-system]
-requires = ["setuptools>=61"]
-build-backend = "setuptools.build_meta"
-
-[project]
-name = "d365fo-agent-developer"
-version = "0.6.0"
-description = "Local D365 F&O knowledge toolkit: corpus indexing (SQLite FTS5), deterministic generation, and an MCP server that grounds coding agents (Claude Code, Codex) in real AOT facts."
-readme = "README.md"
-requires-python = ">=3.11"
-license = { text = "MIT" }
-authors = [{ name = "David Bru", email = "dbru@fiveforty.fr" }]
-keywords = [
-    "dynamics-365", "d365fo", "x++", "aot", "mcp", "model-context-protocol",
-    "claude", "claude-code", "codex", "ai", "code-generation", "rag",
-]
-classifiers = [
-    "Development Status :: 4 - Beta",
-    "Intended Audience :: Developers",
-    "License :: OSI Approved :: MIT License",
-    "Operating System :: OS Independent",
-    "Programming Language :: Python :: 3",
-    "Programming Language :: Python :: 3.11",
-    "Programming Language :: Python :: 3.12",
-    "Programming Language :: Python :: 3.13",
-    "Topic :: Software Development :: Code Generators",
-    "Topic :: Software Development :: Build Tools",
-]
-# Standard library only — nothing to install at runtime, which keeps the MCP server portable.
-dependencies = []
-
-[project.urls]
-Homepage = "https://github.com/dbru540/d365fo-agent"
-Repository = "https://github.com/dbru540/d365fo-agent"
-Issues = "https://github.com/dbru540/d365fo-agent/issues"
-
-[project.optional-dependencies]
-# The graph-building engine is the ONLY external dependency, and only the
-# `run-graphify-staging` command needs it. Everything else (indexing, generation,
-# MCP server, validation) is Python standard library only.
-graph = ["graphify"]
-dev = ["pytest", "ruff", "build"]
-
-[project.scripts]
-d365fo-agent = "d365fo_agent.cli:main"
-d365fo-mcp = "d365fo_agent.mcp_server:main"
-
-[tool.setuptools]
-package-dir = { "" = "src" }
-
-[tool.setuptools.packages.find]
-where = ["src"]
-
-[tool.setuptools.package-data]
-# Methodology, default lint rules, and a default learned type-profile ship inside the package so
-# the server works after a plain `pip install` (no source tree needed).
-d365fo_agent = ["data/*"]
+[build-system]
+requires = ["setuptools>=61"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "d365fo-agent-developer"
+version = "0.6.2"
+description = "Local D365 F&O knowledge toolkit: corpus indexing (SQLite FTS5), deterministic generation, and an MCP server that grounds coding agents (Claude Code, Codex) in real AOT facts."
+readme = "README.md"
+requires-python = ">=3.11"
+license = { text = "MIT" }
+authors = [{ name = "David Bru", email = "dbru@fiveforty.fr" }]
+keywords = [
+    "dynamics-365", "d365fo", "x++", "aot", "mcp", "model-context-protocol",
+    "claude", "claude-code", "codex", "ai", "code-generation", "rag",
+]
+classifiers = [
+    "Development Status :: 4 - Beta",
+    "Intended Audience :: Developers",
+    "License :: OSI Approved :: MIT License",
+    "Operating System :: OS Independent",
+    "Programming Language :: Python :: 3",
+    "Programming Language :: Python :: 3.11",
+    "Programming Language :: Python :: 3.12",
+    "Programming Language :: Python :: 3.13",
+    "Topic :: Software Development :: Code Generators",
+    "Topic :: Software Development :: Build Tools",
+]
+# Standard library only — nothing to install at runtime, which keeps the MCP server portable.
+dependencies = []
+
+[project.urls]
+Homepage = "https://github.com/dbru540/d365fo-agent-developer"
+Repository = "https://github.com/dbru540/d365fo-agent-developer"
+Issues = "https://github.com/dbru540/d365fo-agent-developer/issues"
+
+[project.optional-dependencies]
+# The graph-building engine is the ONLY external dependency, and only the
+# `run-graphify-staging` command needs it. Everything else (indexing, generation,
+# MCP server, validation) is Python standard library only.
+graph = ["graphify"]
+dev = ["pytest", "ruff", "build"]
+
+[project.scripts]
+d365fo-agent = "d365fo_agent.cli:main"
+d365fo-agent-developer = "d365fo_agent.mcp_server:main"
+# Kept as a short alias of d365fo-agent-developer.
+d365fo-mcp = "d365fo_agent.mcp_server:main"
+
+[tool.setuptools]
+package-dir = { "" = "src" }
+
+[tool.setuptools.packages.find]
+where = ["src"]
+
+[tool.setuptools.package-data]
+# Methodology, default lint rules, and a default learned type-profile ship inside the package so
+# the server works after a plain `pip install` (no source tree needed).
+d365fo_agent = ["data/*"]

{d365fo_agent_developer-0.6.0 → d365fo_agent_developer-0.6.2}/src/d365fo_agent/knowledge_fetch.py

@@ -1,70 +1,70 @@
-"""Download a prebuilt standard-D365 knowledge index into the local cache.
-
-This is what lets the MCP server help with X++ development WITHOUT the user re-indexing a full
-repo every time: the standard D365 corpus is the same for everyone, so it is indexed once and
-published as a release asset. ``fetch-knowledge`` downloads it to ``~/.d365fo-agent/d365fo.db``
-(a ``.gz`` asset is decompressed on the fly). The user's own custom code stays optional
-(``--repo-root``). Standard-library only (``urllib`` + ``gzip``).
-"""
-
-from __future__ import annotations
-
-import gzip
-import shutil
-import urllib.request
-from pathlib import Path
-
-# Published location of the standard-D365 knowledge index (a SQLite .db or .db.gz). The asset is
-# decoupled from the code version (tag `knowledge-v1`) so wheel releases don't re-upload 100 MB.
-# Override with --url, or build your own from a PLD when the asset is unavailable.
-DEFAULT_KNOWLEDGE_URL = (
-    "https://github.com/dbru540/d365fo-agent/releases/download/knowledge-v1/d365fo-standard.db.gz"
-)
-
-
-def _is_gzip(path: Path) -> bool:
-    with open(path, "rb") as handle:
-        return handle.read(2) == b"\x1f\x8b"
-
-
-def fetch_knowledge(
-    url: str | None = None,
-    dest: str | Path | None = None,
-    *,
-    force: bool = False,
-    opener=urllib.request.urlopen,
-) -> dict[str, object]:
-    """Download the knowledge index to ``dest`` (default: the local cache). Returns a result dict.
-
-    ``opener`` is injectable for testing. Only http(s) URLs are accepted.
-    """
-    from d365fo_agent.mcp_server import default_knowledge_db
-
-    dest = Path(dest) if dest else default_knowledge_db()
-    url = url or DEFAULT_KNOWLEDGE_URL
-    if not url:
-        return {
-            "ok": False,
-            "error": "No knowledge URL configured. Pass --url <release asset .db/.db.gz>, or build "
-                     "your own with: d365fo-agent build-index --db <out.db> --packages-root <your "
-                     "D365 PackagesLocalDirectory> --rebuild",
-        }
-    if not (url.startswith("http://") or url.startswith("https://")):
-        return {"ok": False, "error": f"Refusing non-http(s) URL: {url}"}
-    if dest.exists() and not force:
-        return {"ok": True, "skipped": True, "dest": str(dest),
-                "note": "knowledge index already present (use --force to re-download)"}
-
-    dest.parent.mkdir(parents=True, exist_ok=True)
-    tmp = dest.with_name(dest.name + ".part")
-    with opener(url) as response, open(tmp, "wb") as out:  # noqa: S310 - scheme checked above
-        shutil.copyfileobj(response, out)
-
-    if url.endswith(".gz") or _is_gzip(tmp):
-        with gzip.open(tmp, "rb") as gz, open(dest, "wb") as out:
-            shutil.copyfileobj(gz, out)
-        tmp.unlink()
-    else:
-        tmp.replace(dest)
-
-    return {"ok": True, "dest": str(dest), "bytes": dest.stat().st_size, "source": url}
+"""Download a prebuilt standard-D365 knowledge index into the local cache.
+
+This is what lets the MCP server help with X++ development WITHOUT the user re-indexing a full
+repo every time: the standard D365 corpus is the same for everyone, so it is indexed once and
+published as a release asset. ``fetch-knowledge`` downloads it to ``~/.d365fo-agent/d365fo.db``
+(a ``.gz`` asset is decompressed on the fly). The user's own custom code stays optional
+(``--repo-root``). Standard-library only (``urllib`` + ``gzip``).
+"""
+
+from __future__ import annotations
+
+import gzip
+import shutil
+import urllib.request
+from pathlib import Path
+
+# Published location of the standard-D365 knowledge index (a SQLite .db or .db.gz). The asset is
+# decoupled from the code version (tag `knowledge-v1`) so wheel releases don't re-upload 100 MB.
+# Override with --url, or build your own from a PLD when the asset is unavailable.
+DEFAULT_KNOWLEDGE_URL = (
+    "https://github.com/dbru540/d365fo-agent-developer/releases/download/knowledge-v1/d365fo-standard.db.gz"
+)
+
+
+def _is_gzip(path: Path) -> bool:
+    with open(path, "rb") as handle:
+        return handle.read(2) == b"\x1f\x8b"
+
+
+def fetch_knowledge(
+    url: str | None = None,
+    dest: str | Path | None = None,
+    *,
+    force: bool = False,
+    opener=urllib.request.urlopen,
+) -> dict[str, object]:
+    """Download the knowledge index to ``dest`` (default: the local cache). Returns a result dict.
+
+    ``opener`` is injectable for testing. Only http(s) URLs are accepted.
+    """
+    from d365fo_agent.mcp_server import default_knowledge_db
+
+    dest = Path(dest) if dest else default_knowledge_db()
+    url = url or DEFAULT_KNOWLEDGE_URL
+    if not url:
+        return {
+            "ok": False,
+            "error": "No knowledge URL configured. Pass --url <release asset .db/.db.gz>, or build "
+                     "your own with: d365fo-agent build-index --db <out.db> --packages-root <your "
+                     "D365 PackagesLocalDirectory> --rebuild",
+        }
+    if not (url.startswith("http://") or url.startswith("https://")):
+        return {"ok": False, "error": f"Refusing non-http(s) URL: {url}"}
+    if dest.exists() and not force:
+        return {"ok": True, "skipped": True, "dest": str(dest),
+                "note": "knowledge index already present (use --force to re-download)"}
+
+    dest.parent.mkdir(parents=True, exist_ok=True)
+    tmp = dest.with_name(dest.name + ".part")
+    with opener(url) as response, open(tmp, "wb") as out:  # noqa: S310 - scheme checked above
+        shutil.copyfileobj(response, out)
+
+    if url.endswith(".gz") or _is_gzip(tmp):
+        with gzip.open(tmp, "rb") as gz, open(dest, "wb") as out:
+            shutil.copyfileobj(gz, out)
+        tmp.unlink()
+    else:
+        tmp.replace(dest)
+
+    return {"ok": True, "dest": str(dest), "bytes": dest.stat().st_size, "source": url}

{d365fo_agent_developer-0.6.0 → d365fo_agent_developer-0.6.2}/src/d365fo_agent/mcp_server.py

@@ -36,13 +36,18 @@ from d365fo_agent.index_store import D365Index
 from d365fo_agent.validate import validate_xml
 
 PROTOCOL_VERSION = "2025-06-18"
-SERVER_NAME = "d365fo-agent"
-SERVER_VERSION = "0.2.0"
+SERVER_NAME = "d365fo-agent-developer"
+try:
+    from importlib.metadata import version as _dist_version
+
+    SERVER_VERSION = _dist_version("d365fo-agent-developer")
+except Exception:  # source tree / not installed
+    SERVER_VERSION = "0.0.0-dev"
 
 
 def _log(message: str) -> None:
     # stdout is the JSON-RPC channel; diagnostics MUST go to stderr.
-    print(f"[d365fo-mcp] {message}", file=sys.stderr, flush=True)
+    print(f"[{SERVER_NAME}] {message}", file=sys.stderr, flush=True)
 
 
 class D365MCPServer:
@@ -799,7 +804,8 @@ def build_server_from_config(
 
 
 def main(argv: list[str] | None = None) -> int:
-    parser = argparse.ArgumentParser(prog="d365fo-mcp", description="D365 F&O knowledge MCP server (stdio).")
+    parser = argparse.ArgumentParser(prog="d365fo-agent-developer",
+                                     description="D365 F&O knowledge MCP server (stdio).")
     parser.add_argument("--repo-root", default=os.environ.get("D365FO_REPO_ROOT"))
     parser.add_argument("--rules", default=os.environ.get("D365FO_RULES"))
     parser.add_argument("--db", default=os.environ.get("D365FO_DB"))

{d365fo_agent_developer-0.6.0 → d365fo_agent_developer-0.6.2/src/d365fo_agent_developer.egg-info}/PKG-INFO

@@ -1,12 +1,12 @@
 Metadata-Version: 2.4
 Name: d365fo-agent-developer
-Version: 0.6.0
+Version: 0.6.2
 Summary: Local D365 F&O knowledge toolkit: corpus indexing (SQLite FTS5), deterministic generation, and an MCP server that grounds coding agents (Claude Code, Codex) in real AOT facts.
 Author-email: David Bru <dbru@fiveforty.fr>
 License: MIT
-Project-URL: Homepage, https://github.com/dbru540/d365fo-agent
-Project-URL: Repository, https://github.com/dbru540/d365fo-agent
-Project-URL: Issues, https://github.com/dbru540/d365fo-agent/issues
+Project-URL: Homepage, https://github.com/dbru540/d365fo-agent-developer
+Project-URL: Repository, https://github.com/dbru540/d365fo-agent-developer
+Project-URL: Issues, https://github.com/dbru540/d365fo-agent-developer/issues
 Keywords: dynamics-365,d365fo,x++,aot,mcp,model-context-protocol,claude,claude-code,codex,ai,code-generation,rag
 Classifier: Development Status :: 4 - Beta
 Classifier: Intended Audience :: Developers
@@ -29,7 +29,7 @@ Requires-Dist: ruff; extra == "dev"
 Requires-Dist: build; extra == "dev"
 Dynamic: license-file
 
-# d365fo-agent — D365 F&O X++ knowledge for Claude Code & Codex
+# d365fo-agent-developer — D365 F&O X++ knowledge for Claude Code & Codex
 
 A local **MCP server** that gives an AI coding agent (Claude Code, Codex, any MCP host) a grounded
 **knowledge base of Dynamics 365 Finance & Operations X++** — so you can develop for D365 quickly
@@ -51,9 +51,14 @@ compiler.
 ```bash
 pip install d365fo-agent-developer
 # or, isolated:  pipx install d365fo-agent-developer
+# or zero-install, straight from PyPI at launch time:  uvx d365fo-agent-developer
 ```
 
-This installs two commands: `d365fo-mcp` (the server) and `d365fo-agent` (the CLI).
+With `uvx` you can even skip the install entirely — point your MCP client at
+`uvx` with args `["d365fo-agent-developer"]` (e.g. in Claude Desktop's
+`claude_desktop_config.json`) and the latest published server runs on demand.
+
+This installs two commands: `d365fo-agent-developer` (the MCP server; `d365fo-mcp` is a short alias) and `d365fo-agent` (the CLI).
 
 ## Get the knowledge base (once)
 
@@ -84,18 +89,18 @@ d365fo-agent build-index \
 ```json
 {
   "mcpServers": {
-    "d365fo": { "command": "d365fo-mcp", "args": [] }
+    "d365fo-agent-developer": { "command": "d365fo-agent-developer", "args": [] }
   }
 }
 ```
 
-…or one command: `claude mcp add d365fo d365fo-mcp`
+…or one command: `claude mcp add d365fo-agent-developer d365fo-agent-developer`
 
 **Codex** — add to `~/.codex/config.toml`:
 
 ```toml
-[mcp_servers.d365fo]
-command = "d365fo-mcp"
+[mcp_servers.d365fo-agent-developer]
+command = "d365fo-agent-developer"
 args = []
 ```
 
@@ -109,8 +114,8 @@ Point the server at your D365 source repo so your **custom** classes/tables/EDTs
 are indexed too, and so the rich tools can read real signatures and clone real examples:
 
 ```toml
-[mcp_servers.d365fo]
-command = "d365fo-mcp"
+[mcp_servers.d365fo-agent-developer]
+command = "d365fo-agent-developer"
 args = ["--repo-root", "C:/path/to/your/D365Repo",
         "--rules", "C:/path/to/your/rules.json",
         "--packages-root", "C:/AOSService/PackagesLocalDirectory"]
@@ -127,7 +132,8 @@ args = ["--repo-root", "C:/path/to/your/D365Repo",
 `element_exists`, `find_element`, `search_corpus`, `get_signature`, `get_extension_chain`,
 `get_security_links`, `get_entity_exposure`, `find_similar_examples`, `scaffold_object`,
 `find_references`, `find_reverse_references`, `analyze_spec`, `generate_from_spec`, `validate_xml`,
-`lint_artifact`, `derive_entity`, `wire_security`, `compile_model`, `get_methodology`, `index_stats`.
+`lint_artifact`, `derive_entity`, `wire_security`, `compile_model`, `compile_generated`,
+`get_sql_model`, `explore_functional_unit`, `find_relations`, `get_methodology`, `index_stats`.
 
 See [docs/mcp-server.md](docs/mcp-server.md) for the verify-driven workflow and
 [docs/x++-methodology.md](docs/x++-methodology.md) for the behavioural contract.

{d365fo_agent_developer-0.6.0 → d365fo_agent_developer-0.6.2}/src/d365fo_agent_developer.egg-info/entry_points.txt

@@ -1,3 +1,4 @@
 [console_scripts]
 d365fo-agent = d365fo_agent.cli:main
+d365fo-agent-developer = d365fo_agent.mcp_server:main
 d365fo-mcp = d365fo_agent.mcp_server:main

{d365fo_agent_developer-0.6.0 → d365fo_agent_developer-0.6.2}/tests/test_mcp_toolkit.py

@@ -463,7 +463,7 @@ class McpServerTests(unittest.TestCase):
 
     def test_initialize(self) -> None:
         resp = self._rpc("initialize", {"protocolVersion": "2025-06-18", "capabilities": {}})
-        self.assertEqual(resp["result"]["serverInfo"]["name"], "d365fo-agent")
+        self.assertEqual(resp["result"]["serverInfo"]["name"], "d365fo-agent-developer")
         self.assertIn("tools", resp["result"]["capabilities"])
 
     def test_tools_list_nonempty(self) -> None: