hexgate 0.2.1__tar.gz → 0.2.3__tar.gz

hexgate-0.2.3/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Hexamind
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

{hexgate-0.2.1/hexgate.egg-info → hexgate-0.2.3}/PKG-INFO

@@ -1,9 +1,11 @@
 Metadata-Version: 2.4
 Name: hexgate
-Version: 0.2.1
-Summary: HexaGate — authorization infrastructure for AI agents (agent runtime + cloud client).
+Version: 0.2.3
+Summary: Hexgate — authorization infrastructure for AI agents (agent runtime + cloud client).
+License-Expression: MIT
 Requires-Python: >=3.13
 Description-Content-Type: text/markdown
+License-File: LICENSE
 Requires-Dist: bashlex>=0.18
 Requires-Dist: biscuit-python>=0.4
 Requires-Dist: cryptography>=42
@@ -33,21 +35,86 @@ Requires-Dist: jupyter; extra == "dev"
 Requires-Dist: pytest>=8.4.1; extra == "dev"
 Requires-Dist: pytest-asyncio>=1.0.0; extra == "dev"
 Requires-Dist: ruff>=0.12.2; extra == "dev"
+Dynamic: license-file
 
-# hexgate
+<div align="center">
 
-`hexgate` is a lightweight LangChain-based agent runtime built around:
+<img src="./icon.svg" alt="Hexgate" width="96" height="96" />
 
-- `langchain`
-- `gpt-5.4`
-- `Linkup` web search
-- Tavily-based page fetch
-- `Langfuse` tracing
+# Hexgate
 
-This package is intentionally small. The first milestone is a single assistant with:
+**Authorization infrastructure for AI agents.**
+Policy enforcement, signed policy bundles, per-request user scope, audit trail — for OpenAI Agents, LangChain, Google ADK, Pydantic AI, or a native runtime.
 
-- `web_search`
-- `fetch`
+[**Website**](https://hexgate.ai) · [Docs](https://docs.hexgate.ai) · [PyPI](https://pypi.org/project/hexgate/) · [Discussions](https://github.com/HexamindOrganisation/hexgate/discussions)
+
+[![PyPI](https://img.shields.io/pypi/v/hexgate?color=blue&logo=pypi&logoColor=white)](https://pypi.org/project/hexgate/)
+[![CI](https://github.com/HexamindOrganisation/hexgate/actions/workflows/tests.yml/badge.svg?branch=main)](https://github.com/HexamindOrganisation/hexgate/actions/workflows/tests.yml)
+[![Downloads](https://img.shields.io/pypi/dm/hexgate?color=blueviolet)](https://pypi.org/project/hexgate/)
+[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](LICENSE)
+
+<br />
+
+<img src="./assets/hero.png" alt="Control what your agents do — not just what they say. Policy decisions streaming live from the PolicyEnforcer." />
+
+<br />
+
+[Quick Start](#-quick-start--local-cli) · [Two paths](#-which-path-do-i-pick) · [Framework adapters](#-framework-agent-wrapping) · [Policy bundles](#-policy-bundles--compile-sign-enforce-wasm) · [User scope](#-user-scope--roles) · [Platform](#-hexgate-platform)
+
+</div>
+
+---
+
+## What is Hexgate?
+
+Hexgate is two things that move together:
+
+- **`hexgate` — the SDK.** A Python runtime that gates every tool call through a typed `Decision` (allow / deny / approval-required), wraps your existing OpenAI / LangChain / Google ADK / Pydantic AI agent without rewriting it, and threads per-request user identity through tracing + audit.
+- **The Hexgate platform** *(optional)* — a FastAPI control plane + React dashboard for editing policy in a browser, minting per-project tokens, watching live decisions stream from a serving agent, and shipping signed WASM policy bundles to production.
+
+You can use the SDK with nothing else (single-process REPL, YAML on disk). Or plug in the platform when you want auditable decisions in ClickHouse, a shared Playground UI, and live policy edits.
+
+```text
+                      ┌─────────────────────────────────────────┐
+   your code  ───►    │   create_agent / wrap_*_agent / Runner  │
+                      │            ↓                            │
+                      │     PolicyEnforcer.decide(role, tool)   │
+                      │            ↓                            │
+                      │   allow · deny · approval_required      │
+                      └────────────────────┬────────────────────┘
+                                           │
+                  ┌────────────────────────┼─────────────────────────┐
+                  ▼                        ▼                         ▼
+        ┌────────────────┐       ┌──────────────────┐       ┌────────────────┐
+        │  Local policy  │       │ Signed WASM      │       │   Audit log    │
+        │  (YAML / dir,  │       │ bundle from      │       │   (ClickHouse  │
+        │  hot reload)   │       │ Hexgate cloud   │       │   via REST)    │
+        └────────────────┘       └──────────────────┘       └────────────────┘
+```
+
+## Table of contents
+
+- [Prerequisites](#-prerequisites)
+- [Quick Start — Local CLI](#-quick-start--local-cli)
+- [Which path do I pick?](#-which-path-do-i-pick) — chat vs serve
+- [Quick Start — Platform](#-quick-start--platform)
+- [Core primitives](#-core-primitives)
+- [Build an agent — end to end](#-build-an-agent--end-to-end)
+- [What you can import](#-what-you-can-import)
+- [Framework agent wrapping](#-framework-agent-wrapping) — OpenAI, LangChain, Google ADK, Pydantic AI
+- [Define agents in code](#-define-agents-in-code)
+- [Builtin and local agents](#-builtin-and-local-agents)
+- [Policy shape](#-policy-shape)
+- [Tool-call policy enforcement](#-tool-call-policy-enforcement)
+- [Policy bundles — compile, sign, enforce (WASM)](#-policy-bundles--compile-sign-enforce-wasm)
+- [Approval-required tool calls](#-approval-required-tool-calls)
+- [Workspace sandbox](#-workspace-sandbox)
+- [Environment](#-environment)
+- [Tests & dev tooling](#-tests--dev-tooling)
+- [CLI reference](#-cli-reference)
+- [Hexgate platform](#-hexgate-platform)
+- [User scope + roles](#-user-scope--roles)
+- [Stream results](#-stream-results)
 
 ## 🛠️ Prerequisites
 
@@ -119,7 +186,7 @@ Both commands accept either a plain agent id (`--agent researcher`) or a uvicorn
 
 ## 🚀 Quick Start — Platform
 
-To run the full HexaGate control plane locally (FastAPI backend + dashboard + your local agent serving over WebSocket), you need **three terminals**. The Makefile has a target that prints the recipe:
+To run the full Hexgate control plane locally (FastAPI backend + dashboard + your local agent serving over WebSocket), you need **three terminals**. The Makefile has a target that prints the recipe:
 
 ```bash
 make demo-platform     # prints the 3-terminal recipe below
@@ -602,7 +669,7 @@ What happens under the hood:
 
 Working scripts in `examples/`:
 
-- `examples/customer_bot.py` — canonical HexaGate path: `create_agent(...)` + the dashboard register/serve loop end-to-end.
+- `examples/customer_bot.py` — canonical Hexgate path: `create_agent(...)` + the dashboard register/serve loop end-to-end.
 - `examples/openai_demo.py` — `HexgateRunner` (OpenAI Agents SDK) end-to-end.
 - `examples/google_demo.py` — `HexgateRunner` (Google ADK) end-to-end with `InMemorySessionService`.
 - `examples/pydantic_ai_demo.py` — `wrap_pydantic_agent` (Pydantic AI) end-to-end.
@@ -737,7 +804,7 @@ That means the same agent code can stay simple in development, while deployment
 
 ## 🧩 Policy Bundles — Compile, Sign, Enforce (WASM)
 
-HexaGate has **two policy enforcement engines** that return identical decisions (there's a parity test suite that proves it):
+Hexgate has **two policy enforcement engines** that return identical decisions (there's a parity test suite that proves it):
 
 - **pydantic** (default) — evaluates constraints in-process. Zero setup; this is what every example above uses.
 - **WASM** — compiles `policy.yaml` → Rego → a WebAssembly module evaluated via `wasmtime`. This is the path production ships: one compiled artifact, byte-for-byte reproducible, cryptographically signed by the platform.
@@ -1059,44 +1126,21 @@ The platform-side test suite is separate and lives at `platform/api/tests/`:
 cd platform/api && uv run pytest tests/
 ```
 
-## ▶️ Run It
-
-Install the package into your current environment:
-
-```bash
-python -m pip install -e .
-```
-
-Run the config-driven demo:
-
-```bash
-python examples/demo.py
-```
-
-Run the inline chat CLI with a local or builtin YAML agent:
-
-```bash
-hexgate chat --agent example_agent
-```
+## 🖥️ CLI reference
 
-Run the CLI with code-defined agents from a Python script:
+The `hexgate` binary exposes `chat`, `serve`, `register`, and `policy` subcommands. The [Quick Start](#-quick-start--local-cli) covers `chat`; this section drills into `register` and `serve` for the platform path.
 
 ```bash
+hexgate --help                                     # list subcommands
+hexgate <subcommand> --help                        # flags for a subcommand
+hexgate chat --list-agents                         # show resolvable agents
 hexgate chat --use examples/file_agents.py --agent workspace_explorer
-hexgate chat --use examples/file_agents.py --agent repo_editor
-hexgate chat --use examples/research_agents.py --agent update_researcher
 hexgate chat --use examples/research_agents.py --agent update_researcher --approval-mode ask
 ```
 
-List what the CLI can currently resolve:
-
-```bash
-hexgate chat --list-agents
-```
-
 ### `hexgate register` — push a manifest to the platform
 
-Register a code-defined agent's manifest with the HexaGate platform. `--agent`
+Register a code-defined agent's manifest with the Hexgate platform. `--agent`
 takes a Python import path of the form `module.path:attribute`, the same shape
 as ASGI/WSGI entrypoints. The CLI imports the module, grabs the agent object,
 and POSTs its manifest to `${HEXGATE_API_URL}/v1/agents` using
@@ -1135,7 +1179,7 @@ system prompt directly off the object. No flags needed.
 `--system-prompt` accepts either a literal string or a path to a `.md` /
 `.txt` / `.jinja` file (read as text at register time).
 
-Supported frameworks: OpenAI Agents SDK, Google ADK, Pydantic AI, LangChain/LangGraph, HexaGate agents.
+Supported frameworks: OpenAI Agents SDK, Google ADK, Pydantic AI, LangChain/LangGraph, Hexgate agents.
 
 ### `hexgate serve` — bridge a local agent to the platform's relay
 
@@ -1172,7 +1216,7 @@ print(manifest.model_dump())
 ```
 
 `create_manifest` dispatches on the framework of `agent`. The supported
-types are the same set `hexgate register` accepts: HexaGate, OpenAI Agents
+types are the same set `hexgate register` accepts: Hexgate, OpenAI Agents
 SDK, Google ADK, Pydantic AI, and LangChain/LangGraph compiled graphs.
 For LangGraph you must pass `tools=` explicitly, and may pass `model=` /
 `system_prompt=`, since compiled graphs don't expose those fields after
@@ -1182,7 +1226,7 @@ The return value is an `AgentManifest` (a Pydantic model, also re-exported
 from `hexgate` for type annotations) — the same schema the platform
 stores and the dashboard renders.
 
-## 🌐 HexaGate Platform
+## 🌐 Hexgate Platform
 
 The `platform/` directory contains an optional control plane that hosts agent definitions, dev tokens, and a live debug surface. The SDK works fully without it (`load_local_agent`, `load_builtin_agent` keep their existing semantics) — but with it you get:
 
@@ -1301,7 +1345,7 @@ the name from the loaded agent's `.name` attribute — no env var needed.
 
 ## 👤 User Scope + Roles
 
-Real backends serve many users, and different users get different capabilities. HexaGate splits that into two pieces:
+Real backends serve many users, and different users get different capabilities. Hexgate splits that into two pieces:
 
 - **`User`** — the per-request scope. Marks "this invocation acts on behalf of alice, in role X." Async context manager; pushes a fact-bearing Biscuit through the agent runtime.
 - **Role policies** — one `policy.yaml` per role, optionally inheriting from a base mixin. The runtime picks the right one at call time based on the active `User.role`.
@@ -1451,3 +1495,7 @@ async for event in stream_agent(agent, handler, "latest AI breakthroughs"):
 - assistant text deltas
 - tool lifecycle
 - final run completion
+
+---
+
+If Hexgate looks useful, [give it a ⭐ on GitHub](https://github.com/HexamindOrganisation/hexgate) — it helps more than you'd think. Built by [Hexamind](https://hexgate.ai).

hexgate-0.2.1/PKG-INFO → hexgate-0.2.3/README.md

@@ -1,53 +1,81 @@
-Metadata-Version: 2.4
-Name: hexgate
-Version: 0.2.1
-Summary: HexaGate — authorization infrastructure for AI agents (agent runtime + cloud client).
-Requires-Python: >=3.13
-Description-Content-Type: text/markdown
-Requires-Dist: bashlex>=0.18
-Requires-Dist: biscuit-python>=0.4
-Requires-Dist: cryptography>=42
-Requires-Dist: httpx>=0.28.1
-Requires-Dist: langchain
-Requires-Dist: langchain-openai
-Requires-Dist: langchain-core
-Requires-Dist: langfuse
-Requires-Dist: pydantic>=2.12.4
-Requires-Dist: python-dotenv>=1.1.1
-Requires-Dist: pyyaml>=6.0.2
-Requires-Dist: rich>=13.9.4
-Requires-Dist: websockets>=13.0
-Requires-Dist: openai-agents>=0.0.10
-Requires-Dist: langgraph>=0.2
-Requires-Dist: nest_asyncio>=1.6
-Requires-Dist: openinference-instrumentation-openai-agents>=0.1
-Requires-Dist: google-adk>=1.0
-Requires-Dist: google-genai>=1.0
-Requires-Dist: litellm>=1.50
-Requires-Dist: openinference-instrumentation-google-adk>=0.1.11
-Requires-Dist: pydantic-ai-slim>=1.88.0
-Requires-Dist: wasmtime>=20.0
-Provides-Extra: dev
-Requires-Dist: ipykernel; extra == "dev"
-Requires-Dist: jupyter; extra == "dev"
-Requires-Dist: pytest>=8.4.1; extra == "dev"
-Requires-Dist: pytest-asyncio>=1.0.0; extra == "dev"
-Requires-Dist: ruff>=0.12.2; extra == "dev"
-
-# hexgate
-
-`hexgate` is a lightweight LangChain-based agent runtime built around:
-
-- `langchain`
-- `gpt-5.4`
-- `Linkup` web search
-- Tavily-based page fetch
-- `Langfuse` tracing
-
-This package is intentionally small. The first milestone is a single assistant with:
+<div align="center">
 
-- `web_search`
-- `fetch`
+<img src="./icon.svg" alt="Hexgate" width="96" height="96" />
+
+# Hexgate
+
+**Authorization infrastructure for AI agents.**
+Policy enforcement, signed policy bundles, per-request user scope, audit trail — for OpenAI Agents, LangChain, Google ADK, Pydantic AI, or a native runtime.
+
+[**Website**](https://hexgate.ai) · [Docs](https://docs.hexgate.ai) · [PyPI](https://pypi.org/project/hexgate/) · [Discussions](https://github.com/HexamindOrganisation/hexgate/discussions)
+
+[![PyPI](https://img.shields.io/pypi/v/hexgate?color=blue&logo=pypi&logoColor=white)](https://pypi.org/project/hexgate/)
+[![CI](https://github.com/HexamindOrganisation/hexgate/actions/workflows/tests.yml/badge.svg?branch=main)](https://github.com/HexamindOrganisation/hexgate/actions/workflows/tests.yml)
+[![Downloads](https://img.shields.io/pypi/dm/hexgate?color=blueviolet)](https://pypi.org/project/hexgate/)
+[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](LICENSE)
+
+<br />
+
+<img src="./assets/hero.png" alt="Control what your agents do — not just what they say. Policy decisions streaming live from the PolicyEnforcer." />
+
+<br />
+
+[Quick Start](#-quick-start--local-cli) · [Two paths](#-which-path-do-i-pick) · [Framework adapters](#-framework-agent-wrapping) · [Policy bundles](#-policy-bundles--compile-sign-enforce-wasm) · [User scope](#-user-scope--roles) · [Platform](#-hexgate-platform)
+
+</div>
+
+---
+
+## What is Hexgate?
+
+Hexgate is two things that move together:
+
+- **`hexgate` — the SDK.** A Python runtime that gates every tool call through a typed `Decision` (allow / deny / approval-required), wraps your existing OpenAI / LangChain / Google ADK / Pydantic AI agent without rewriting it, and threads per-request user identity through tracing + audit.
+- **The Hexgate platform** *(optional)* — a FastAPI control plane + React dashboard for editing policy in a browser, minting per-project tokens, watching live decisions stream from a serving agent, and shipping signed WASM policy bundles to production.
+
+You can use the SDK with nothing else (single-process REPL, YAML on disk). Or plug in the platform when you want auditable decisions in ClickHouse, a shared Playground UI, and live policy edits.
+
+```text
+                      ┌─────────────────────────────────────────┐
+   your code  ───►    │   create_agent / wrap_*_agent / Runner  │
+                      │            ↓                            │
+                      │     PolicyEnforcer.decide(role, tool)   │
+                      │            ↓                            │
+                      │   allow · deny · approval_required      │
+                      └────────────────────┬────────────────────┘
+                                           │
+                  ┌────────────────────────┼─────────────────────────┐
+                  ▼                        ▼                         ▼
+        ┌────────────────┐       ┌──────────────────┐       ┌────────────────┐
+        │  Local policy  │       │ Signed WASM      │       │   Audit log    │
+        │  (YAML / dir,  │       │ bundle from      │       │   (ClickHouse  │
+        │  hot reload)   │       │ Hexgate cloud   │       │   via REST)    │
+        └────────────────┘       └──────────────────┘       └────────────────┘
+```
+
+## Table of contents
+
+- [Prerequisites](#-prerequisites)
+- [Quick Start — Local CLI](#-quick-start--local-cli)
+- [Which path do I pick?](#-which-path-do-i-pick) — chat vs serve
+- [Quick Start — Platform](#-quick-start--platform)
+- [Core primitives](#-core-primitives)
+- [Build an agent — end to end](#-build-an-agent--end-to-end)
+- [What you can import](#-what-you-can-import)
+- [Framework agent wrapping](#-framework-agent-wrapping) — OpenAI, LangChain, Google ADK, Pydantic AI
+- [Define agents in code](#-define-agents-in-code)
+- [Builtin and local agents](#-builtin-and-local-agents)
+- [Policy shape](#-policy-shape)
+- [Tool-call policy enforcement](#-tool-call-policy-enforcement)
+- [Policy bundles — compile, sign, enforce (WASM)](#-policy-bundles--compile-sign-enforce-wasm)
+- [Approval-required tool calls](#-approval-required-tool-calls)
+- [Workspace sandbox](#-workspace-sandbox)
+- [Environment](#-environment)
+- [Tests & dev tooling](#-tests--dev-tooling)
+- [CLI reference](#-cli-reference)
+- [Hexgate platform](#-hexgate-platform)
+- [User scope + roles](#-user-scope--roles)
+- [Stream results](#-stream-results)
 
 ## 🛠️ Prerequisites
 
@@ -119,7 +147,7 @@ Both commands accept either a plain agent id (`--agent researcher`) or a uvicorn
 
 ## 🚀 Quick Start — Platform
 
-To run the full HexaGate control plane locally (FastAPI backend + dashboard + your local agent serving over WebSocket), you need **three terminals**. The Makefile has a target that prints the recipe:
+To run the full Hexgate control plane locally (FastAPI backend + dashboard + your local agent serving over WebSocket), you need **three terminals**. The Makefile has a target that prints the recipe:
 
 ```bash
 make demo-platform     # prints the 3-terminal recipe below
@@ -602,7 +630,7 @@ What happens under the hood:
 
 Working scripts in `examples/`:
 
-- `examples/customer_bot.py` — canonical HexaGate path: `create_agent(...)` + the dashboard register/serve loop end-to-end.
+- `examples/customer_bot.py` — canonical Hexgate path: `create_agent(...)` + the dashboard register/serve loop end-to-end.
 - `examples/openai_demo.py` — `HexgateRunner` (OpenAI Agents SDK) end-to-end.
 - `examples/google_demo.py` — `HexgateRunner` (Google ADK) end-to-end with `InMemorySessionService`.
 - `examples/pydantic_ai_demo.py` — `wrap_pydantic_agent` (Pydantic AI) end-to-end.
@@ -737,7 +765,7 @@ That means the same agent code can stay simple in development, while deployment
 
 ## 🧩 Policy Bundles — Compile, Sign, Enforce (WASM)
 
-HexaGate has **two policy enforcement engines** that return identical decisions (there's a parity test suite that proves it):
+Hexgate has **two policy enforcement engines** that return identical decisions (there's a parity test suite that proves it):
 
 - **pydantic** (default) — evaluates constraints in-process. Zero setup; this is what every example above uses.
 - **WASM** — compiles `policy.yaml` → Rego → a WebAssembly module evaluated via `wasmtime`. This is the path production ships: one compiled artifact, byte-for-byte reproducible, cryptographically signed by the platform.
@@ -1059,44 +1087,21 @@ The platform-side test suite is separate and lives at `platform/api/tests/`:
 cd platform/api && uv run pytest tests/
 ```
 
-## ▶️ Run It
+## 🖥️ CLI reference
 
-Install the package into your current environment:
-
-```bash
-python -m pip install -e .
-```
-
-Run the config-driven demo:
-
-```bash
-python examples/demo.py
-```
-
-Run the inline chat CLI with a local or builtin YAML agent:
-
-```bash
-hexgate chat --agent example_agent
-```
-
-Run the CLI with code-defined agents from a Python script:
+The `hexgate` binary exposes `chat`, `serve`, `register`, and `policy` subcommands. The [Quick Start](#-quick-start--local-cli) covers `chat`; this section drills into `register` and `serve` for the platform path.
 
 ```bash
+hexgate --help                                     # list subcommands
+hexgate <subcommand> --help                        # flags for a subcommand
+hexgate chat --list-agents                         # show resolvable agents
 hexgate chat --use examples/file_agents.py --agent workspace_explorer
-hexgate chat --use examples/file_agents.py --agent repo_editor
-hexgate chat --use examples/research_agents.py --agent update_researcher
 hexgate chat --use examples/research_agents.py --agent update_researcher --approval-mode ask
 ```
 
-List what the CLI can currently resolve:
-
-```bash
-hexgate chat --list-agents
-```
-
 ### `hexgate register` — push a manifest to the platform
 
-Register a code-defined agent's manifest with the HexaGate platform. `--agent`
+Register a code-defined agent's manifest with the Hexgate platform. `--agent`
 takes a Python import path of the form `module.path:attribute`, the same shape
 as ASGI/WSGI entrypoints. The CLI imports the module, grabs the agent object,
 and POSTs its manifest to `${HEXGATE_API_URL}/v1/agents` using
@@ -1135,7 +1140,7 @@ system prompt directly off the object. No flags needed.
 `--system-prompt` accepts either a literal string or a path to a `.md` /
 `.txt` / `.jinja` file (read as text at register time).
 
-Supported frameworks: OpenAI Agents SDK, Google ADK, Pydantic AI, LangChain/LangGraph, HexaGate agents.
+Supported frameworks: OpenAI Agents SDK, Google ADK, Pydantic AI, LangChain/LangGraph, Hexgate agents.
 
 ### `hexgate serve` — bridge a local agent to the platform's relay
 
@@ -1172,7 +1177,7 @@ print(manifest.model_dump())
 ```
 
 `create_manifest` dispatches on the framework of `agent`. The supported
-types are the same set `hexgate register` accepts: HexaGate, OpenAI Agents
+types are the same set `hexgate register` accepts: Hexgate, OpenAI Agents
 SDK, Google ADK, Pydantic AI, and LangChain/LangGraph compiled graphs.
 For LangGraph you must pass `tools=` explicitly, and may pass `model=` /
 `system_prompt=`, since compiled graphs don't expose those fields after
@@ -1182,7 +1187,7 @@ The return value is an `AgentManifest` (a Pydantic model, also re-exported
 from `hexgate` for type annotations) — the same schema the platform
 stores and the dashboard renders.
 
-## 🌐 HexaGate Platform
+## 🌐 Hexgate Platform
 
 The `platform/` directory contains an optional control plane that hosts agent definitions, dev tokens, and a live debug surface. The SDK works fully without it (`load_local_agent`, `load_builtin_agent` keep their existing semantics) — but with it you get:
 
@@ -1301,7 +1306,7 @@ the name from the loaded agent's `.name` attribute — no env var needed.
 
 ## 👤 User Scope + Roles
 
-Real backends serve many users, and different users get different capabilities. HexaGate splits that into two pieces:
+Real backends serve many users, and different users get different capabilities. Hexgate splits that into two pieces:
 
 - **`User`** — the per-request scope. Marks "this invocation acts on behalf of alice, in role X." Async context manager; pushes a fact-bearing Biscuit through the agent runtime.
 - **Role policies** — one `policy.yaml` per role, optionally inheriting from a base mixin. The runtime picks the right one at call time based on the active `User.role`.
@@ -1451,3 +1456,7 @@ async for event in stream_agent(agent, handler, "latest AI breakthroughs"):
 - assistant text deltas
 - tool lifecycle
 - final run completion
+
+---
+
+If Hexgate looks useful, [give it a ⭐ on GitHub](https://github.com/HexamindOrganisation/hexgate) — it helps more than you'd think. Built by [Hexamind](https://hexgate.ai).

{hexgate-0.2.1 → hexgate-0.2.3}/hexgate/adapters/google/runner.py

@@ -21,7 +21,7 @@ from hexgate.runtime import User
 
 
 class HexgateRunner:
-    """Runner for Google ADK agents with HexaGate tool policy and observability."""
+    """Runner for Google ADK agents with Hexgate tool policy and observability."""
 
     def __init__(
         self,
@@ -79,16 +79,32 @@ class HexgateRunner:
         user: User,
         **kwargs: Any,
     ) -> Generator[Any, None, None]:
-        """Run the Google ADK agent synchronously, yielding events."""
+        """Run the Google ADK agent synchronously, yielding events.
+
+        ADK's ``Runner.run`` drives the agent loop in a worker thread whose
+        context cannot see our :class:`User` scope, so the tools' enforcers
+        lose the active role. We drive ``run_async`` inline on a per-call loop
+        instead, keeping execution in this scoped thread.
+        """
         self._setup_observability()
         self._binding.refresh()  # per-run policy pull; 304 when unchanged
         with user.sync_scope(), self._propagate(user):
-            yield from self._runner.run(
+            agen = self._runner.run_async(
                 user_id=user.user_id,
                 session_id=user.session_id,
                 new_message=new_message,
                 **kwargs,
             )
+            loop = asyncio.new_event_loop()
+            try:
+                while True:
+                    try:
+                        yield loop.run_until_complete(agen.__anext__())
+                    except StopAsyncIteration:
+                        break
+            finally:
+                loop.run_until_complete(agen.aclose())
+                loop.close()
 
     async def run_async(
         self,

{hexgate-0.2.1 → hexgate-0.2.3}/hexgate/adapters/langchain/agent.py

@@ -1,4 +1,4 @@
-"""Proxy around a pre-built ``CompiledStateGraph`` for HexaGate-aware calls."""
+"""Proxy around a pre-built ``CompiledStateGraph`` for Hexgate-aware calls."""
 
 from __future__ import annotations
 
@@ -60,7 +60,7 @@ class HexgateLangchainAgent:
         }
 
     def _with_callbacks(self, config: RunnableConfig | None) -> RunnableConfig:
-        """Append the HexaGate callback handler to ``config['callbacks']``."""
+        """Append the Hexgate callback handler to ``config['callbacks']``."""
         merged: RunnableConfig = dict(config) if config else {}
         callbacks = list(merged.get("callbacks") or [])
         if self._callback_handler not in callbacks:

{hexgate-0.2.1 → hexgate-0.2.3}/hexgate/adapters/langchain/wrapper.py

@@ -1,5 +1,5 @@
 """BYO-graph entry point: retrofit a pre-built ``CompiledStateGraph`` with
-HexaGate policy. Tools are mutated in place so the graph keeps its
+Hexgate policy. Tools are mutated in place so the graph keeps its
 references; the returned :class:`HexgateLangchainAgent` opens a User
 scope + Langfuse propagation per call. For the manifest-driven path,
 use :func:`hexgate.enforce_policy` instead.
@@ -28,7 +28,7 @@ def wrap_langchain_agent(
     tools: list[BaseTool],
     api_key: str | None = None,
 ) -> HexgateLangchainAgent:
-    """Wrap a pre-built LangGraph agent with HexaGate policy enforcement.
+    """Wrap a pre-built LangGraph agent with Hexgate policy enforcement.
 
     Mutates ``tools`` in place so the graph keeps its references.
     The returned proxy takes ``user`` per invocation; role resolves at

{hexgate-0.2.1 → hexgate-0.2.3}/hexgate/adapters/openai/runner.py

@@ -32,7 +32,7 @@ from hexgate.security.enforcer import build_enforcer
 
 
 class HexgateRunner:
-    """Runner for OpenAI agents with HexaGate tool policy and observability."""
+    """Runner for OpenAI agents with Hexgate tool policy and observability."""
 
     def __init__(self, api_key: str | None = None):
         self.api_key = api_key or os.getenv("HEXGATE_KEY")
@@ -134,21 +134,22 @@ class HexgateRunner:
     ) -> RunResultStreaming:
         """Stream the OpenAI agent inside a User scope.
 
-        ``Runner.run_streamed`` returns sync; tools only run during
-        ``stream_events`` iteration. The User scope is opened inside the
-        wrapped iterator. Langfuse propagation runs during setup so the
-        trace span attaches. The policy refresh runs in the setup body —
-        the wrap is fixed before tools fire during stream_events.
+        ``Runner.run_streamed`` returns sync but spawns the agent loop as a
+        background task that snapshots the current contextvars at creation;
+        tools fire there, not in ``stream_events``. So the User scope must be
+        active around the ``run_streamed`` call for the task to inherit it —
+        the wrapped iterator re-opens it for exit/audit semantics.
         """
         self._setup_observability()
         binding = self._binding_for(agent)
         binding.refresh()  # must precede the wrap + setup
         wrapped_agent = wrap_openai_agent(agent, enforcer=binding.enforcer)
 
-        with self._propagate(user, agent.name):
-            result = Runner.run_streamed(
-                wrapped_agent, input, run_config=run_config, **kwargs
-            )
+        with user.sync_scope():
+            with self._propagate(user, agent.name):
+                result = Runner.run_streamed(
+                    wrapped_agent, input, run_config=run_config, **kwargs
+                )
 
         original_stream_events = result.stream_events
 

{hexgate-0.2.1 → hexgate-0.2.3}/hexgate/adapters/pydantic_ai/agent.py

@@ -1,4 +1,4 @@
-"""Proxy around a pydantic_ai ``Agent`` for HexaGate-aware calls."""
+"""Proxy around a pydantic_ai ``Agent`` for Hexgate-aware calls."""
 
 from __future__ import annotations
 

{hexgate-0.2.1 → hexgate-0.2.3}/hexgate/adapters/pydantic_ai/wrapper.py

@@ -49,7 +49,7 @@ def wrap_pydantic_agent(
     agent: Agent,
     api_key: str | None = None,
 ) -> HexgatePydanticAgent:
-    """Wrap a pydantic_ai agent with HexaGate policy + observability.
+    """Wrap a pydantic_ai agent with Hexgate policy + observability.
 
     Returns a :class:`HexgatePydanticAgent` backed by a clone of the
     caller's ``agent``; the original is not mutated. The proxy takes