enbanc 0.0.1__tar.gz

enbanc-0.0.1/LICENSE

@@ -0,0 +1,19 @@
+Copyright (c) 2026 Andy Tseng
+
+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.

enbanc-0.0.1/PKG-INFO

@@ -0,0 +1,99 @@
+Metadata-Version: 2.4
+Name: enbanc
+Version: 0.0.1
+Summary: Adversarial multi-agent adjudication for structured decisions
+Author-email: Andy Tseng <ohandyya@gmail.com>
+License-Expression: MIT
+Project-URL: Homepage, https://github.com/ohandyya/enbanc
+Project-URL: Repository, https://github.com/ohandyya/enbanc
+Keywords: llm,agents,multi-agent,adjudication,pydantic-ai
+Classifier: Development Status :: 2 - Pre-Alpha
+Classifier: Intended Audience :: Developers
+Classifier: Programming Language :: Python :: 3
+Classifier: Typing :: Typed
+Requires-Python: >=3.11
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Dynamic: license-file
+
+# enbanc
+
+**Adversarial multi-agent adjudication.** Advocates argue each possible answer;
+a judge cross-examines them until it can rule — with a full audit transcript.
+
+Built on [PydanticAI](https://ai.pydantic.dev).
+
+## Why
+
+Most "LLM as judge" tooling asks one model to score an output. `enbanc` instead
+stages an adversarial proceeding: every possible answer gets a dedicated
+advocate with its own tools, and a tool-less judge interrogates them across
+rounds until it can rule or the round limit is hit.
+
+That structure is designed for decisions that need to be **defensible** —
+loan underwriting, insurance appetite, eligibility screening — where you need
+the reasoning and the evidence, not just the answer.
+
+## Install
+
+```bash
+uv add enbanc
+```
+
+## Usage
+
+```python
+from enbanc import Tribunal, Advocate, Statute, Case, Verdict
+
+class LoanDecision(Verdict):
+    APPROVE = "approve"
+    DENY = "deny"
+
+statute = Statute.draft(
+    "Approve $500k loans only where DTI < 0.43 and ...",
+    model="anthropic:claude-sonnet-4-6",
+)
+
+tribunal = Tribunal(
+    question="Shall the bank loan this applicant $500k?",
+    verdicts=LoanDecision,
+    statute=statute,
+    advocates={
+        LoanDecision.APPROVE: Advocate(tools=[psql, tavily]),
+        LoanDecision.DENY: Advocate(tools=[psql]),
+    },
+    max_rounds=5,
+)
+
+ruling = await tribunal.hear(Case(applicant=..., income=...))
+
+ruling.verdict      # LoanDecision.DENY
+ruling.reasoning
+ruling.transcript   # every argument, exhibit, and interrogatory, in order
+```
+
+## How it works
+
+**Round 1** — each advocate argues freely for its assigned verdict, or concedes
+that no reasonable case exists. The judge hears all arguments and evidence.
+
+**Round 2+** — if the judge cannot rule, it issues *interrogatories*: targeted
+questions to a chosen subset of advocates. Those advocates answer. Repeat.
+
+Ends when the judge rules, or `max_rounds` is exceeded.
+
+The judge has **no tools** — it reasons only over what advocates put into the
+record. All advocate tools are **strictly read-only**.
+
+## Vocabulary
+
+`enbanc` borrows courtroom terms because they're precise. See
+[`glossary.md`](./glossary.md) — six rows, one minute.
+
+## Status
+
+Early. API will change.
+
+## License
+
+MIT

enbanc-0.0.1/README.md

@@ -0,0 +1,81 @@
+# enbanc
+
+**Adversarial multi-agent adjudication.** Advocates argue each possible answer;
+a judge cross-examines them until it can rule — with a full audit transcript.
+
+Built on [PydanticAI](https://ai.pydantic.dev).
+
+## Why
+
+Most "LLM as judge" tooling asks one model to score an output. `enbanc` instead
+stages an adversarial proceeding: every possible answer gets a dedicated
+advocate with its own tools, and a tool-less judge interrogates them across
+rounds until it can rule or the round limit is hit.
+
+That structure is designed for decisions that need to be **defensible** —
+loan underwriting, insurance appetite, eligibility screening — where you need
+the reasoning and the evidence, not just the answer.
+
+## Install
+
+```bash
+uv add enbanc
+```
+
+## Usage
+
+```python
+from enbanc import Tribunal, Advocate, Statute, Case, Verdict
+
+class LoanDecision(Verdict):
+    APPROVE = "approve"
+    DENY = "deny"
+
+statute = Statute.draft(
+    "Approve $500k loans only where DTI < 0.43 and ...",
+    model="anthropic:claude-sonnet-4-6",
+)
+
+tribunal = Tribunal(
+    question="Shall the bank loan this applicant $500k?",
+    verdicts=LoanDecision,
+    statute=statute,
+    advocates={
+        LoanDecision.APPROVE: Advocate(tools=[psql, tavily]),
+        LoanDecision.DENY: Advocate(tools=[psql]),
+    },
+    max_rounds=5,
+)
+
+ruling = await tribunal.hear(Case(applicant=..., income=...))
+
+ruling.verdict      # LoanDecision.DENY
+ruling.reasoning
+ruling.transcript   # every argument, exhibit, and interrogatory, in order
+```
+
+## How it works
+
+**Round 1** — each advocate argues freely for its assigned verdict, or concedes
+that no reasonable case exists. The judge hears all arguments and evidence.
+
+**Round 2+** — if the judge cannot rule, it issues *interrogatories*: targeted
+questions to a chosen subset of advocates. Those advocates answer. Repeat.
+
+Ends when the judge rules, or `max_rounds` is exceeded.
+
+The judge has **no tools** — it reasons only over what advocates put into the
+record. All advocate tools are **strictly read-only**.
+
+## Vocabulary
+
+`enbanc` borrows courtroom terms because they're precise. See
+[`glossary.md`](./glossary.md) — six rows, one minute.
+
+## Status
+
+Early. API will change.
+
+## License
+
+MIT

enbanc-0.0.1/pyproject.toml

@@ -0,0 +1,21 @@
+[project]
+name = "enbanc"
+version = "0.0.1"
+description = "Adversarial multi-agent adjudication for structured decisions"
+readme = "README.md"
+requires-python = ">=3.11"
+license = "MIT"
+license-files = ["LICENSE"]
+authors = [{ name = "Andy Tseng", email = "ohandyya@gmail.com" }]
+keywords = ["llm", "agents", "multi-agent", "adjudication", "pydantic-ai"]
+classifiers = [
+    "Development Status :: 2 - Pre-Alpha",
+    "Intended Audience :: Developers",
+    "Programming Language :: Python :: 3",
+    "Typing :: Typed",
+]
+dependencies = []
+
+[project.urls]
+Homepage = "https://github.com/ohandyya/enbanc"
+Repository = "https://github.com/ohandyya/enbanc"

enbanc-0.0.1/setup.cfg

@@ -0,0 +1,4 @@
+[egg_info]
+tag_build = 
+tag_date = 0
+

enbanc-0.0.1/src/enbanc/__init__.py

@@ -0,0 +1,2 @@
+def hello() -> str:
+    return "Hello from enbanc!"

enbanc-0.0.1/src/enbanc.egg-info/PKG-INFO

@@ -0,0 +1,99 @@
+Metadata-Version: 2.4
+Name: enbanc
+Version: 0.0.1
+Summary: Adversarial multi-agent adjudication for structured decisions
+Author-email: Andy Tseng <ohandyya@gmail.com>
+License-Expression: MIT
+Project-URL: Homepage, https://github.com/ohandyya/enbanc
+Project-URL: Repository, https://github.com/ohandyya/enbanc
+Keywords: llm,agents,multi-agent,adjudication,pydantic-ai
+Classifier: Development Status :: 2 - Pre-Alpha
+Classifier: Intended Audience :: Developers
+Classifier: Programming Language :: Python :: 3
+Classifier: Typing :: Typed
+Requires-Python: >=3.11
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Dynamic: license-file
+
+# enbanc
+
+**Adversarial multi-agent adjudication.** Advocates argue each possible answer;
+a judge cross-examines them until it can rule — with a full audit transcript.
+
+Built on [PydanticAI](https://ai.pydantic.dev).
+
+## Why
+
+Most "LLM as judge" tooling asks one model to score an output. `enbanc` instead
+stages an adversarial proceeding: every possible answer gets a dedicated
+advocate with its own tools, and a tool-less judge interrogates them across
+rounds until it can rule or the round limit is hit.
+
+That structure is designed for decisions that need to be **defensible** —
+loan underwriting, insurance appetite, eligibility screening — where you need
+the reasoning and the evidence, not just the answer.
+
+## Install
+
+```bash
+uv add enbanc
+```
+
+## Usage
+
+```python
+from enbanc import Tribunal, Advocate, Statute, Case, Verdict
+
+class LoanDecision(Verdict):
+    APPROVE = "approve"
+    DENY = "deny"
+
+statute = Statute.draft(
+    "Approve $500k loans only where DTI < 0.43 and ...",
+    model="anthropic:claude-sonnet-4-6",
+)
+
+tribunal = Tribunal(
+    question="Shall the bank loan this applicant $500k?",
+    verdicts=LoanDecision,
+    statute=statute,
+    advocates={
+        LoanDecision.APPROVE: Advocate(tools=[psql, tavily]),
+        LoanDecision.DENY: Advocate(tools=[psql]),
+    },
+    max_rounds=5,
+)
+
+ruling = await tribunal.hear(Case(applicant=..., income=...))
+
+ruling.verdict      # LoanDecision.DENY
+ruling.reasoning
+ruling.transcript   # every argument, exhibit, and interrogatory, in order
+```
+
+## How it works
+
+**Round 1** — each advocate argues freely for its assigned verdict, or concedes
+that no reasonable case exists. The judge hears all arguments and evidence.
+
+**Round 2+** — if the judge cannot rule, it issues *interrogatories*: targeted
+questions to a chosen subset of advocates. Those advocates answer. Repeat.
+
+Ends when the judge rules, or `max_rounds` is exceeded.
+
+The judge has **no tools** — it reasons only over what advocates put into the
+record. All advocate tools are **strictly read-only**.
+
+## Vocabulary
+
+`enbanc` borrows courtroom terms because they're precise. See
+[`glossary.md`](./glossary.md) — six rows, one minute.
+
+## Status
+
+Early. API will change.
+
+## License
+
+MIT

enbanc-0.0.1/src/enbanc.egg-info/SOURCES.txt

@@ -0,0 +1,9 @@
+LICENSE
+README.md
+pyproject.toml
+src/enbanc/__init__.py
+src/enbanc/py.typed
+src/enbanc.egg-info/PKG-INFO
+src/enbanc.egg-info/SOURCES.txt
+src/enbanc.egg-info/dependency_links.txt
+src/enbanc.egg-info/top_level.txt

enbanc-0.0.1/src/enbanc.egg-info/dependency_links.txt

@@ -0,0 +1 @@
+

enbanc-0.0.1/src/enbanc.egg-info/top_level.txt

@@ -0,0 +1 @@
+enbanc