@yahyahammad/venastine 0.3.11 → 0.3.12
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +52 -3
- package/package.json +19 -2
- package/pyproject.toml +15 -1
- package/requirements.txt +8 -8
package/README.md
CHANGED
|
@@ -1,7 +1,22 @@
|
|
|
1
|
-

|
|
2
|
-
|
|
3
1
|
# Venastine Research Harness
|
|
4
2
|
|
|
3
|
+
**Venastine Research Harness is an open-source, security-first agentic harness for large language models: a tool-using chat assistant and a ten-pass deep-research pipeline that fact-checks, critiques and scores its own claims, in your terminal or in the browser.**
|
|
4
|
+
|
|
5
|
+
<p align="center">
|
|
6
|
+
<a href="https://huggingface.co/spaces/SolenopsisCampo/VenastineHarnessDemo"><img alt="Try the demo on a Hugging Face Space" src="https://img.shields.io/badge/Demo-Hugging%20Face%20Space-1783FF?logo=huggingface&logoColor=white"></a>
|
|
7
|
+
<a href="https://www.modelscope.ai/studios/Venastine-Research/VenastineHarnessDemo"><img alt="Try the demo on a ModelScope Studio" src="https://img.shields.io/badge/Demo-ModelScope%20Studio-1783FF?logo=modelscope&logoColor=white"></a>
|
|
8
|
+
<a href="https://huggingface.co/Venastine-Research"><img alt="Venastine-Research on Hugging Face" src="https://img.shields.io/badge/Hugging%20Face-Venastine--Research-FFC107?logo=huggingface&logoColor=white"></a>
|
|
9
|
+
<a href="https://www.modelscope.ai/organization/Venastine-Research"><img alt="Venastine-Research on ModelScope" src="https://img.shields.io/badge/ModelScope-Venastine--Research-624AFF?logo=modelscope&logoColor=white"></a>
|
|
10
|
+
<br>
|
|
11
|
+
<a href="https://github.com/YahyaHammad/Venastine-Research-Harness/releases/latest"><img alt="Latest GitHub release" src="https://img.shields.io/github/v/release/YahyaHammad/Venastine-Research-Harness?logo=github&label=release"></a>
|
|
12
|
+
<a href="https://www.npmjs.com/package/@yahyahammad/venastine"><img alt="npm package @yahyahammad/venastine" src="https://img.shields.io/npm/v/%40yahyahammad%2Fvenastine?logo=npm&label=npm"></a>
|
|
13
|
+
<a href="https://github.com/YahyaHammad/Venastine-Research-Harness/actions/workflows/tests.yml"><img alt="Tests on main" src="https://img.shields.io/github/actions/workflow/status/YahyaHammad/Venastine-Research-Harness/tests.yml?branch=main&label=tests&logo=githubactions&logoColor=white"></a>
|
|
14
|
+
<a href="#setup"><img alt="Python 3.11, 3.12 and 3.13" src="https://img.shields.io/badge/python-3.11%20%7C%203.12%20%7C%203.13-3776AB?logo=python&logoColor=white"></a>
|
|
15
|
+
<a href="./LICENSE"><img alt="License: Apache-2.0" src="https://img.shields.io/badge/license-Apache--2.0-F5DE53"></a>
|
|
16
|
+
</p>
|
|
17
|
+
|
|
18
|
+

|
|
19
|
+
|
|
5
20
|
An agentic harness built from scratch — no LangChain, no agent framework — with two modes that share the same machinery:
|
|
6
21
|
|
|
7
22
|
- **Chat** — a normal tool-using assistant loop.
|
|
@@ -122,6 +137,40 @@ Two things differ from a clone, both to make a command you run from anywhere beh
|
|
|
122
137
|
|
|
123
138
|
`venastine --venastine-doctor` prints every one of those resolved paths, plus the interpreter and environment it found.
|
|
124
139
|
|
|
140
|
+
#### With yarn, bun or pnpm
|
|
141
|
+
|
|
142
|
+
The same package installs from the npm registry with the other package
|
|
143
|
+
managers, and everything above about the launcher applies to each. The
|
|
144
|
+
launcher still needs Node.js 18 or newer, and the harness Python 3.11 or newer.
|
|
145
|
+
|
|
146
|
+
```bash
|
|
147
|
+
yarn global add @yahyahammad/venastine
|
|
148
|
+
```
|
|
149
|
+
|
|
150
|
+
```bash
|
|
151
|
+
bun add -g --backend=copyfile @yahyahammad/venastine
|
|
152
|
+
```
|
|
153
|
+
|
|
154
|
+
```bash
|
|
155
|
+
pnpm add -g --config.package-import-method=copy @yahyahammad/venastine
|
|
156
|
+
```
|
|
157
|
+
|
|
158
|
+
To update, run the same command with `@latest` after the package name.
|
|
159
|
+
|
|
160
|
+
* **yarn means Yarn 1 (classic).** Yarn 2 and later have no global install.
|
|
161
|
+
* **bun and pnpm need the copy flag, on updates too.** By default both
|
|
162
|
+
hardlink the package's files out of a shared cache or store, so the
|
|
163
|
+
harness's own source, which it protects from tool writes, would have a
|
|
164
|
+
second name in a folder it does not protect. The flag makes them copy,
|
|
165
|
+
as npm and yarn do.
|
|
166
|
+
* **pnpm does not carry `config.yaml` edits across an update.** It puts
|
|
167
|
+
every install in a new directory, and the harness keys your saved
|
|
168
|
+
`config.yaml` values by the install's directory, so after an update the
|
|
169
|
+
values you set with `/config` or by editing that file are back at the
|
|
170
|
+
shipped defaults. What lives in `~/.config/venastine` (`settings.json`,
|
|
171
|
+
keys, conversations and memories) is unaffected. If you rely on
|
|
172
|
+
`config.yaml` edits, install with npm, yarn or bun instead.
|
|
173
|
+
|
|
125
174
|
### From a clone
|
|
126
175
|
|
|
127
176
|
```bash
|
|
@@ -1015,7 +1064,7 @@ note that shell approval is governed solely by `shell_approval_mode` -- `tool_ap
|
|
|
1015
1064
|
deliberately has no `shell` key, so an old file carrying one is refused at startup
|
|
1016
1065
|
naming the key — see `tests/BREAKING_CHANGES.md` §24.
|
|
1017
1066
|
|
|
1018
|
-
Run the test suite with `pytest` —
|
|
1067
|
+
Run the test suite with `pytest` — 9199 tests, fully offline, no API keys needed. Since batch 243 the suite holds itself to that: a test that asks DNS for a name, or writes into your home folder, fails by name, and every test process runs with HOME pointing at a throwaway folder of its own. The suite's own packages are `requirements-test.txt`, separate from `requirements.txt` since batch 136: they used to be in the shipped dependency list, so installing the harness also installed a test runner. It runs in parallel by default (`pytest-xdist`, 70% of your cores): **1:26 on a 16-core machine, against 9:06 single-process.** `pytest -n0` forces the old serial run, which is what you want under a debugger or when something is parsing the output. Eight further tests are marked `integration` and excluded by default: three spawn a real stdio MCP server, one runs an external program through the host fallback shell, one builds the hosted bundle's Docker image, and three run the installers' uv path against a real uv (`pytest -m integration`).
|
|
1019
1068
|
|
|
1020
1069
|
## Documentation
|
|
1021
1070
|
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@yahyahammad/venastine",
|
|
3
|
-
"version": "0.3.
|
|
3
|
+
"version": "0.3.12",
|
|
4
4
|
"description": "A from-scratch agentic harness: tool-using chat plus a ten-pass self-critiquing research pipeline.",
|
|
5
5
|
"keywords": [
|
|
6
6
|
"agent",
|
|
@@ -8,7 +8,24 @@
|
|
|
8
8
|
"research",
|
|
9
9
|
"anthropic",
|
|
10
10
|
"openai",
|
|
11
|
-
"mcp"
|
|
11
|
+
"mcp",
|
|
12
|
+
"agentic-ai",
|
|
13
|
+
"ai-agent",
|
|
14
|
+
"agent-harness",
|
|
15
|
+
"llm-agent",
|
|
16
|
+
"research-agent",
|
|
17
|
+
"deep-research",
|
|
18
|
+
"fact-checking",
|
|
19
|
+
"tool-use",
|
|
20
|
+
"model-context-protocol",
|
|
21
|
+
"gemini",
|
|
22
|
+
"deepseek",
|
|
23
|
+
"openrouter",
|
|
24
|
+
"hugging-face",
|
|
25
|
+
"modelscope",
|
|
26
|
+
"security",
|
|
27
|
+
"cli",
|
|
28
|
+
"tui"
|
|
12
29
|
],
|
|
13
30
|
"homepage": "https://github.com/YahyaHammad/Venastine-Research-Harness#readme",
|
|
14
31
|
"bugs": {
|
package/pyproject.toml
CHANGED
|
@@ -18,9 +18,17 @@ build-backend = "setuptools.build_meta"
|
|
|
18
18
|
|
|
19
19
|
[project]
|
|
20
20
|
name = "venastine-research-harness"
|
|
21
|
-
version = "0.3.
|
|
21
|
+
version = "0.3.12"
|
|
22
22
|
description = "A from-scratch agentic harness: tool-using chat plus a ten-pass self-critiquing research pipeline."
|
|
23
23
|
readme = "README.md"
|
|
24
|
+
# The same words as package.json's `keywords`, in the same order: what a
|
|
25
|
+
# package index searches. test_docs_consistency.py holds the two lists equal.
|
|
26
|
+
keywords = [
|
|
27
|
+
"agent", "llm", "research", "anthropic", "openai", "mcp", "agentic-ai",
|
|
28
|
+
"ai-agent", "agent-harness", "llm-agent", "research-agent", "deep-research",
|
|
29
|
+
"fact-checking", "tool-use", "model-context-protocol", "gemini", "deepseek",
|
|
30
|
+
"openrouter", "hugging-face", "modelscope", "security", "cli", "tui",
|
|
31
|
+
]
|
|
24
32
|
requires-python = ">=3.11"
|
|
25
33
|
# PEP 639: a bare SPDX expression, not the deprecated `{ text = ... }` table, and no
|
|
26
34
|
# `License :: OSI Approved ::` classifier -- PEP 639 supersedes those and setuptools warns on
|
|
@@ -39,6 +47,9 @@ classifiers = [
|
|
|
39
47
|
"Programming Language :: Python :: 3.11",
|
|
40
48
|
"Programming Language :: Python :: 3.12",
|
|
41
49
|
"Programming Language :: Python :: 3.13",
|
|
50
|
+
"Environment :: Console",
|
|
51
|
+
"Operating System :: OS Independent",
|
|
52
|
+
"Intended Audience :: Developers",
|
|
42
53
|
"Intended Audience :: Science/Research",
|
|
43
54
|
"Topic :: Scientific/Engineering :: Artificial Intelligence",
|
|
44
55
|
]
|
|
@@ -63,6 +74,9 @@ dynamic = ["dependencies"]
|
|
|
63
74
|
Homepage = "https://github.com/YahyaHammad/Venastine-Research-Harness"
|
|
64
75
|
Repository = "https://github.com/YahyaHammad/Venastine-Research-Harness"
|
|
65
76
|
Issues = "https://github.com/YahyaHammad/Venastine-Research-Harness/issues"
|
|
77
|
+
Documentation = "https://github.com/YahyaHammad/Venastine-Research-Harness#readme"
|
|
78
|
+
Changelog = "https://github.com/YahyaHammad/Venastine-Research-Harness/releases"
|
|
79
|
+
Demo = "https://huggingface.co/spaces/SolenopsisCampo/VenastineHarnessDemo"
|
|
66
80
|
|
|
67
81
|
[project.optional-dependencies]
|
|
68
82
|
# markitdown's only consumer is tools/builtin/file_ops.py, which implements
|
package/requirements.txt
CHANGED
|
@@ -21,10 +21,10 @@
|
|
|
21
21
|
# test_sdk_conformance.py re-checks the shapes. What to re-measure when
|
|
22
22
|
# these move: that the stream, the effort query and the error types still
|
|
23
23
|
# match, since the suite fakes both packages everywhere else.
|
|
24
|
-
anthropic==2.
|
|
25
|
-
ddgs==9.16.
|
|
24
|
+
anthropic==2.5.0
|
|
25
|
+
ddgs==9.16.1
|
|
26
26
|
httpx==0.28.1
|
|
27
|
-
openai==3.
|
|
27
|
+
openai==3.45.0
|
|
28
28
|
# ROADMAP_v3 §82 (GS1), batch 172: 1.0.0 -> 2.29.0, the owner's choice.
|
|
29
29
|
# Gemini 3 refuses a request whose current turn holds a function call
|
|
30
30
|
# without the thought signature it came with; 1.0.0's Part had no field
|
|
@@ -37,7 +37,7 @@ openai==3.36.0
|
|
|
37
37
|
# transport is httpx now, not requests. 2.32.0 since batch 234, the newest
|
|
38
38
|
# release at least two weeks old; test_sdk_conformance.py is the evidence,
|
|
39
39
|
# with no live Gemini call (no key on this machine).
|
|
40
|
-
google-genai==2.
|
|
40
|
+
google-genai==2.34.0
|
|
41
41
|
# Batch 208, from Dependabot's group, the owner's choice of which: the
|
|
42
42
|
# packages below are ones the suite runs for real rather than faking, so
|
|
43
43
|
# a green suite is evidence about them, and each release was at least two
|
|
@@ -52,7 +52,7 @@ google-genai==2.32.0
|
|
|
52
52
|
# core/reasoning/output_writer.py draws the chart as SVG with the standard
|
|
53
53
|
# library now. pillow's floor (batch 201) went with it: nothing else here
|
|
54
54
|
# reaches pillow.
|
|
55
|
-
pydantic==2.14.
|
|
55
|
+
pydantic==2.14.2
|
|
56
56
|
sympy==1.14.0
|
|
57
57
|
# 0.0.48 since batch 234: test_storage_e2e.py runs it for real.
|
|
58
58
|
sqlmodel==0.0.48
|
|
@@ -122,7 +122,7 @@ textual==8.2.8
|
|
|
122
122
|
# without declaring, i.e. the one place such a migration would be silent.
|
|
123
123
|
# The FLOOR is textual's, not a guess: 8.2.8 requires `rich>=14.2.0`, so
|
|
124
124
|
# `>=13` stated a version this project can no longer be installed with.
|
|
125
|
-
rich>=
|
|
125
|
+
rich>=15.0.0,<16
|
|
126
126
|
|
|
127
127
|
# ROADMAP_v2 §17 MCP client. Upper-bounded per D22 -- and D22 exists
|
|
128
128
|
# BECAUSE of this package: it was mid-major-version transition when §17
|
|
@@ -239,7 +239,7 @@ google-re2==1.1.20251105
|
|
|
239
239
|
# message (the store relies on the three failure causes being
|
|
240
240
|
# indistinguishable), and the wheel list still covers this project's
|
|
241
241
|
# platforms.
|
|
242
|
-
cryptography>=50.0.
|
|
242
|
+
cryptography>=50.0.2,<52
|
|
243
243
|
|
|
244
244
|
# ROADMAP_v3 §83 (CG6, CG16). Checks the ID token a ChatGPT-plan sign-in
|
|
245
245
|
# returns against OpenAI's published keys: RS256, issuer, audience, expiry
|
|
@@ -266,7 +266,7 @@ cryptography>=50.0.0,<52
|
|
|
266
266
|
# moves: `jwt.decode` still refuses an algorithm outside the list it is
|
|
267
267
|
# given (the `none` and HS256 tests in tests/test_chatgpt_auth.py), and
|
|
268
268
|
# `PyJWKSet` still reads OpenAI's JWKS.
|
|
269
|
-
PyJWT>=2.15.
|
|
269
|
+
PyJWT>=2.15.1,<3
|
|
270
270
|
|
|
271
271
|
# TECHNICAL_DEBT 18, batch 116. A PRODUCTION dependency: `arxiv.py` imports
|
|
272
272
|
# it at module load, and pyproject.toml reads this file for its dependency
|