ne-agent 0.1.0__tar.gz

ne_agent-0.1.0/PKG-INFO

@@ -0,0 +1,79 @@
+Metadata-Version: 2.4
+Name: ne-agent
+Version: 0.1.0
+Summary: First open-source AI agent for Northeast Indian languages
+Home-page: https://github.com/MWirelabs/ne-agent
+Author: MWire Labs
+Author-email: connect@mwirelabs.com
+Classifier: Programming Language :: Python :: 3
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Operating System :: OS Independent
+Requires-Python: >=3.9
+Description-Content-Type: text/markdown
+Requires-Dist: ne-lid
+Requires-Dist: ne-embed
+Requires-Dist: faiss-cpu
+Requires-Dist: rich
+Requires-Dist: requests
+Requires-Dist: pandas
+Requires-Dist: openpyxl
+Dynamic: author
+Dynamic: author-email
+Dynamic: classifier
+Dynamic: description
+Dynamic: description-content-type
+Dynamic: home-page
+Dynamic: requires-dist
+Dynamic: requires-python
+Dynamic: summary
+
+# NE-Agent
+
+**First open-source AI agent for Northeast Indian languages.**
+
+Built on the NE-Stack by [MWire Labs](https://mwirelabs.com), Shillong.
+
+---
+
+## What is NE-Agent?
+
+NE-Agent is a terminal-based agentic assistant that speaks Northeast India's indigenous languages. It automatically detects the input language, retrieves relevant context from a multilingual corpus, and generates responses — all running locally with no API keys required.
+
+---
+
+## Powered by NE-Stack
+
+| Component | Role | Model |
+|---|---|---|
+| NE-LID | Language identification | fastText, 11 languages, 99.09% accuracy |
+| NE-Embed | Multilingual embeddings | LaBSE fine-tuned, 768-dim |
+| FAISS | Vector retrieval | IndexFlatIP |
+| Ollama | Local LLM | qwen2.5:1.5b (default) |
+
+---
+
+## Supported Languages
+
+Assamese · Khasi · Garo · Mizo · Meitei · Bodo · Kokborok · Nyishi · Nagamese · English
+
+---
+
+## Quickstart
+
+```bash
+pip install ne-agent
+```
+
+Add your corpus to `data/` and run:
+
+```bash
+python run.py
+```
+
+Type in any Northeast Indian language. Type `exit` or `:q` to quit.
+
+---
+
+## License
+
+CC-BY-4.0 — MWire Labs, Shillong, Meghalaya.

ne_agent-0.1.0/README.md

@@ -0,0 +1,50 @@
+# NE-Agent
+
+**First open-source AI agent for Northeast Indian languages.**
+
+Built on the NE-Stack by [MWire Labs](https://mwirelabs.com), Shillong.
+
+---
+
+## What is NE-Agent?
+
+NE-Agent is a terminal-based agentic assistant that speaks Northeast India's indigenous languages. It automatically detects the input language, retrieves relevant context from a multilingual corpus, and generates responses — all running locally with no API keys required.
+
+---
+
+## Powered by NE-Stack
+
+| Component | Role | Model |
+|---|---|---|
+| NE-LID | Language identification | fastText, 11 languages, 99.09% accuracy |
+| NE-Embed | Multilingual embeddings | LaBSE fine-tuned, 768-dim |
+| FAISS | Vector retrieval | IndexFlatIP |
+| Ollama | Local LLM | qwen2.5:1.5b (default) |
+
+---
+
+## Supported Languages
+
+Assamese · Khasi · Garo · Mizo · Meitei · Bodo · Kokborok · Nyishi · Nagamese · English
+
+---
+
+## Quickstart
+
+```bash
+pip install ne-agent
+```
+
+Add your corpus to `data/` and run:
+
+```bash
+python run.py
+```
+
+Type in any Northeast Indian language. Type `exit` or `:q` to quit.
+
+---
+
+## License
+
+CC-BY-4.0 — MWire Labs, Shillong, Meghalaya.

ne_agent-0.1.0/ne_agent/__init__.py

@@ -0,0 +1,4 @@
+from .core import NEAgent
+
+__version__ = "0.1.0"
+__all__ = ["NEAgent"]

ne_agent-0.1.0/ne_agent/core.py

@@ -0,0 +1,24 @@
+from .lid import LanguageIdentifier
+from .retriever import Retriever
+from .llm import OllamaLLM
+
+class NEAgent:
+    def __init__(self, model: str = "qwen2.5:1.5b"):
+        self.lid = LanguageIdentifier()
+        self.retriever = Retriever()
+        self.llm = OllamaLLM(model=model)
+
+    def load_corpus(self, corpus: dict, n: int = 500):
+        self.retriever.build(corpus, n=n)
+
+    def run(self, query: str) -> dict:
+        lang, score = self.lid.predict(query)
+        docs = self.retriever.retrieve(query, lang, top_k=3)
+        answer = self.llm.generate(query, docs)
+        return {
+            "query": query,
+            "detected_lang": lang,
+            "lang_score": score,
+            "retrieved": docs,
+            "answer": answer
+        }

ne_agent-0.1.0/ne_agent/lid.py

@@ -0,0 +1,9 @@
+from ne_lid import NELID
+
+class LanguageIdentifier:
+    def __init__(self):
+        self.model = NELID()
+
+    def predict(self, text: str) -> tuple[str, float]:
+        result = self.model.predict(text)
+        return result["lang"], round(result["score"], 4)

ne_agent-0.1.0/ne_agent/llm.py

@@ -0,0 +1,24 @@
+import requests
+
+class OllamaLLM:
+    def __init__(self, model: str = "qwen2.5:1.5b", host: str = "http://127.0.0.1:11434"):
+        self.model = model
+        self.host = host
+
+    def generate(self, query: str, context_docs: list) -> str:
+        context = "\n".join([f"[{d['lang']}] {d['text']}" for d in context_docs])
+        prompt = f"""You are NE-Agent, an AI assistant for Northeast Indian languages.
+Use the following retrieved context to answer the query.
+
+Context:
+{context}
+
+Query: {query}
+
+Answer:"""
+        response = requests.post(f"{self.host}/api/generate", json={
+            "model": self.model,
+            "prompt": prompt,
+            "stream": False
+        })
+        return response.json()["response"]

ne_agent-0.1.0/ne_agent/retriever.py

@@ -0,0 +1,38 @@
+import faiss
+import numpy as np
+from ne_embed import NEEmbed
+
+class Retriever:
+    def __init__(self):
+        self.model = NEEmbed()
+        self.index = None
+        self.sentences = []
+        self.langs = []
+
+    def build(self, corpus: dict, n: int = 500):
+        import random
+        random.seed(42)
+        for lang, sents in corpus.items():
+            cleaned = [s.strip() for s in sents if isinstance(s, str) and len(s.strip()) > 10]
+            sampled = random.sample(cleaned, min(n, len(cleaned)))
+            self.sentences.extend(sampled)
+            self.langs.extend([lang] * len(sampled))
+        embeddings = self.model.encode(self.sentences, batch_size=64).astype(np.float32)
+        self.index = faiss.IndexFlatIP(embeddings.shape[1])
+        self.index.add(embeddings)
+        print(f"Index built: {self.index.ntotal} vectors")
+
+    def retrieve(self, query: str, lang: str, top_k: int = 3) -> list:
+        qe = self.model.encode([query]).astype(np.float32)
+        scores, indices = self.index.search(qe, 50)
+        results = []
+        for score, idx in zip(scores[0], indices[0]):
+            if self.langs[idx] == lang:
+                results.append({
+                    "text": self.sentences[idx],
+                    "lang": self.langs[idx],
+                    "score": round(float(score), 4)
+                })
+            if len(results) == top_k:
+                break
+        return results

ne_agent-0.1.0/ne_agent/tui.py

@@ -0,0 +1,69 @@
+import time
+from rich.console import Console
+from rich.text import Text
+from rich.align import Align
+from rich.panel import Panel
+from rich.rule import Rule
+
+console = Console()
+
+GREETINGS = [
+    ("Kumno ngi tip u ia u?", "khasi"),
+    ("Angni manda?", "garo"),
+    ("Eng nge nia?", "mizo"),
+    ("Nongshitla?", "meitei"),
+    ("Kemon acho?", "assamese"),
+    ("Nangse ema?", "bodo"),
+]
+
+def load_banner() -> str:
+    try:
+        with open("assets/banner.txt") as f:
+            return f.read()
+    except:
+        return "NE-AGENT"
+
+def boot_screen():
+    console.clear()
+    console.print(Align.center(Text(load_banner(), style="bold orange1")))
+    console.print(Align.center(Text("Speaks Northeast.", style="bold white")))
+    console.print()
+    for phrase, lang in GREETINGS:
+        line = Text()
+        line.append(f"{phrase}", style="white")
+        line.append(f"  [{lang}]", style="bold blue")
+        console.print(Align.center(line))
+        time.sleep(0.2)
+    console.print()
+    console.print(Rule(style="orange1"))
+    console.print()
+
+def print_status(index_size: int, model: str):
+    status = Text()
+    status.append("● NE-LID  ", style="orange1")
+    status.append("● NE-Embed  ", style="orange1")
+    status.append(f"● Ollama {model}  ", style="orange1")
+    status.append(f"● FAISS index: {index_size} docs", style="orange1")
+    console.print(Align.center(status))
+    console.print()
+    console.print(Rule(style="orange1"))
+    console.print()
+
+def print_result(result: dict):
+    lang_line = Text()
+    lang_line.append("LANG  ", style="dim white")
+    lang_line.append(result["detected_lang"].upper(), style="bold blue")
+    lang_line.append(f"  {result['lang_score']}", style="dim white")
+    console.print(lang_line)
+    console.print()
+    if result["retrieved"]:
+        console.print(Text("RETRIEVED", style="dim orange1"))
+        for d in result["retrieved"]:
+            console.print(Text(f"  · {d['text'][:70]}", style="dim white"))
+    console.print()
+    console.print(Panel(
+        Text(result["answer"], style="white"),
+        border_style="orange1",
+        title=Text("NE-AGENT", style="bold orange1"),
+    ))
+    console.print()

ne_agent-0.1.0/ne_agent.egg-info/PKG-INFO

@@ -0,0 +1,79 @@
+Metadata-Version: 2.4
+Name: ne-agent
+Version: 0.1.0
+Summary: First open-source AI agent for Northeast Indian languages
+Home-page: https://github.com/MWirelabs/ne-agent
+Author: MWire Labs
+Author-email: connect@mwirelabs.com
+Classifier: Programming Language :: Python :: 3
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Operating System :: OS Independent
+Requires-Python: >=3.9
+Description-Content-Type: text/markdown
+Requires-Dist: ne-lid
+Requires-Dist: ne-embed
+Requires-Dist: faiss-cpu
+Requires-Dist: rich
+Requires-Dist: requests
+Requires-Dist: pandas
+Requires-Dist: openpyxl
+Dynamic: author
+Dynamic: author-email
+Dynamic: classifier
+Dynamic: description
+Dynamic: description-content-type
+Dynamic: home-page
+Dynamic: requires-dist
+Dynamic: requires-python
+Dynamic: summary
+
+# NE-Agent
+
+**First open-source AI agent for Northeast Indian languages.**
+
+Built on the NE-Stack by [MWire Labs](https://mwirelabs.com), Shillong.
+
+---
+
+## What is NE-Agent?
+
+NE-Agent is a terminal-based agentic assistant that speaks Northeast India's indigenous languages. It automatically detects the input language, retrieves relevant context from a multilingual corpus, and generates responses — all running locally with no API keys required.
+
+---
+
+## Powered by NE-Stack
+
+| Component | Role | Model |
+|---|---|---|
+| NE-LID | Language identification | fastText, 11 languages, 99.09% accuracy |
+| NE-Embed | Multilingual embeddings | LaBSE fine-tuned, 768-dim |
+| FAISS | Vector retrieval | IndexFlatIP |
+| Ollama | Local LLM | qwen2.5:1.5b (default) |
+
+---
+
+## Supported Languages
+
+Assamese · Khasi · Garo · Mizo · Meitei · Bodo · Kokborok · Nyishi · Nagamese · English
+
+---
+
+## Quickstart
+
+```bash
+pip install ne-agent
+```
+
+Add your corpus to `data/` and run:
+
+```bash
+python run.py
+```
+
+Type in any Northeast Indian language. Type `exit` or `:q` to quit.
+
+---
+
+## License
+
+CC-BY-4.0 — MWire Labs, Shillong, Meghalaya.

ne_agent-0.1.0/ne_agent.egg-info/SOURCES.txt

@@ -0,0 +1,13 @@
+README.md
+setup.py
+ne_agent/__init__.py
+ne_agent/core.py
+ne_agent/lid.py
+ne_agent/llm.py
+ne_agent/retriever.py
+ne_agent/tui.py
+ne_agent.egg-info/PKG-INFO
+ne_agent.egg-info/SOURCES.txt
+ne_agent.egg-info/dependency_links.txt
+ne_agent.egg-info/requires.txt
+ne_agent.egg-info/top_level.txt

ne_agent-0.1.0/ne_agent.egg-info/dependency_links.txt

@@ -0,0 +1 @@
+

ne_agent-0.1.0/ne_agent.egg-info/requires.txt

@@ -0,0 +1,7 @@
+ne-lid
+ne-embed
+faiss-cpu
+rich
+requests
+pandas
+openpyxl

ne_agent-0.1.0/ne_agent.egg-info/top_level.txt

@@ -0,0 +1 @@
+ne_agent

ne_agent-0.1.0/setup.cfg

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

ne_agent-0.1.0/setup.py

@@ -0,0 +1,20 @@
+from setuptools import setup, find_packages
+
+setup(
+    name="ne-agent",
+    version="0.1.0",
+    author="MWire Labs",
+    author_email="connect@mwirelabs.com",
+    description="First open-source AI agent for Northeast Indian languages",
+    long_description=open("README.md").read(),
+    long_description_content_type="text/markdown",
+    url="https://github.com/MWirelabs/ne-agent",
+    packages=find_packages(),
+    install_requires=open("requirements.txt").read().splitlines(),
+    python_requires=">=3.9",
+    classifiers=[
+        "Programming Language :: Python :: 3",
+        "License :: OSI Approved :: MIT License",
+        "Operating System :: OS Independent",
+    ],
+)