0.1.0 is here Python 3.11

Your agent.
Your boundaries.

The runtime beneath your agent application. Typed tools, explicit approvals, and durable sessions—under your application’s control.

Open source · MIT OR Apache-2.0 · Async Python

01 / THE EXECUTION BOUNDARYAPPLICATION OWNED
A TOOL CALL IS A PROPOSAL↘
01Check policyCurrent actor. Current rules.
02Ask when requiredApprove the executable arguments.
03Commit intentRecord before dispatch.
04Run the toolThen commit its result.↗
Uncertain effect? Report it. Don’t silently replay it.
THE MODEL PROPOSES. YOUR APPLICATION AUTHORIZES.
Built on Pydantic AI AgentZero provider SDKs in baseLocal durable JSONLOptional PostgreSQL 18
01 / THE LIBRARY

Bring your interface.
Keep your storage and policy.

The useful work is yours.
The execution details have a home.

01

Tools with boundaries

Typed inputs, pinned definitions, validation, and approval gates. A model response does not grant permission to execute.

02

Facts that survive a restart

Durable execution records and original history have distinct roles. Reopen accepted state, surface unknown effects, and resume explicitly.

03

One API for your interface

An in-process async Client API for session metadata, paged history, persisted changes, and supported commands. Build your own UI around it.

02 / IN PRACTICEACTUAL REFERENCE CLIENT · FABRICATED DATA

Close the app.
Keep the conversation.

A recorded, offline journey through the installed 0.1.0 terminal UI. A tool waits for approval, the turn completes, and the same local store is reopened.

CAPTURE PROFILE

Python 3.11.17 · macOS arm64
Scripted model · network denied

View capture receipt ↗
tether / offline demo01—04
Actual Tether terminal UI showing a fresh offline demo session.
01 / A fresh durable offline demo session.

These are screenshots exported from the real Textual UI, driven through its test pilot with state assertions. This walkthrough is a capture, not an interactive agent running in your browser.

ALSO CAPTURED / LIVE LOCAL INFERENCE

Same interface.
A local Qwen.

These prompts ran against an already-running local DS4 server with qwen3.8-flash-next-chat. The model returned text, proposed a fabricated tool call, and waited for explicit approval.

Installed 0.1.0 · Python 3.11.17 · macOS arm64
384 output tokens · 128 KiB response cap
Fabricated prompts · explicit loopback endpoint

View live-model capture receipt ↗
tether / local QwenLIVE MODEL CAPTURE
Actual Tether UI with a live local Qwen tool proposal waiting for approval.
02 / A live local model proposal is held for approval.
SOURCE PREVIEW / COMMON TOOLS

Review the arguments.
Then approve.

The opt-in reference profile adds bash, local file read/write, web search and page loading. This actual UI capture uses a scripted model and a real fabricated file write. The file exists only after approval.

Unreleased working tree · fabricated data
Bash runs with user permissions; the workspace is not a sandbox.
Live model and search behavior are outside this capture.

View the approved state ↗
View New Session ↗
View capture receipt ↗
Actual Tether UI showing write_file arguments for seedling.txt before approval.
Fabricated file write held for approval; no file exists yet.
SOURCE PREVIEW / URL-ONLY LOCAL STARTUP

Point at your model.
Approve real tools.

The local launcher discovers the served chat model and enables common tools. This actual Qwen journey runs all five tools in a temporary workspace, with approval before each operation.

uv run tether --openai-api-baseurl=http://localhost:18001/v1

Unreleased source · actual local inference
Fabricated prompts and file data · public web query/page
Approximate byte estimates · explicit loopback endpoint

View all five tool checks ↗
Actual Tether local-model TUI requesting approval for a fabricated file write.
A real local-model tool proposal waits for approval before creating the file.
SOURCE PREVIEW / PROFILE-ONLY STARTUP

Choose a profile.
Pick a model.

The Databricks launcher derives its gateway URL and authentication from an existing CLI profile. This actual TUI journey discovers 59 system.ai chat models, selects a public model and verifies a fabricated READY response.

uv run tether --databricks-profile=work

Unreleased source · actual gateway inference
CLI OAuth user profile · macOS / Python 3.11.17
Catalog metadata does not guarantee invocation permission.

View the accepted response ↗
View capture receipt ↗
Actual Tether TUI offering discovered public Databricks chat models before opening a session.
No URL supplied: the selected CLI profile provides the gateway connection.
SOURCE PREVIEW / BROWSER CLIENT

One session.
Your choice of interface.

The optional browser application consumes the same public client API through HTTP. This real Chromium capture shows an exact fabricated tool proposal waiting for keyboard approval, followed by committed history after reloading.

Unreleased development source · fabricated model and tools
Nested UI/API mount · desktop and mobile checks
Separate transport, React hooks and replaceable visual components

View committed history ↗
View mobile screenshot ↗
Read capture scope and recipe ↗
Actual Tether browser UI showing the executable arguments of a fabricated tool call before approval.
Development source screenshot. Approval is explicit; untrusted message markup remains text.
03 / GET STARTED

Start local.
Make it yours.

Install the library for your application, or launch the offline reference client. The demo needs no provider account or model credential.

Read the developer guide ↗
PYTHON 3.11 ENVIRONMENT
uv pip install 'novirhiza-tether==0.1.0'

Import as tether. Provider SDKs are excluded from the base installation.

Set up a local model profile ↗

novirhiza-tether on PyPI ↗
04 / THE SUPPORTED SUBSET

Claims with receipts.

Support follows the tested versions and profiles. The release evidence records passing checks alongside failed attempts and limits.

Read the qualification receipt ↗
Runtime
Python 3.11.17 · Pydantic AI 2.59.0
Platforms
macOS arm64 & Debian 12 Linux arm64 installed profiles. Complete final source gate on Linux.
Persistence
Local POSIX JSONL; qualified PostgreSQL 18. Embedded helper requires an explicit platform binary.
Models
Controlled compatible-HTTP checks and a recorded local DS4 text/tool profile. Other models need their own qualification.
Release
Immutable GitHub artifacts; public PyPI files match the release manifest.
⨁

A small foundation.
Your application on top.

Explore the source ↗