SandyVeliz/acervo-extractor-qwen3.5-9b

Acervo Extractor — Qwen3.5-9B Fine-Tuned

A fine-tuned version of Qwen3.5-9B specialized in knowledge graph extraction from conversations. Given a conversation turn and existing graph context, the model outputs structured JSON with topic classification, entities, relations, and facts.

Built for Acervo — a semantic compression layer for AI agents that replaces raw conversation history with compressed knowledge graph nodes.

What it does

Input: A conversation message + existing graph nodes as context

Output: Structured JSON with:

• Topic classification — same / subtopic / changed
• Entities — people, projects, technologies, events, places, etc.
• Relations — uses_technology, maintains, part_of, participated_in, etc.
• Facts — specific claims attached to existing entities

Example

Input:

EXISTING NODES:
[{"id": "beacon", "label": "Beacon", "type": "project", "layer": "PERSONAL"}]

TOPIC HINT: same (high confidence from keyword match)
CURRENT TOPIC: Beacon development

PREVIOUS ASSISTANT: How's the project going?
USER: Beacon ya tiene 50 mil usuarios y estamos migrando a Kubernetes.

Output:

{
  "topic": {"action": "same"},
  "entities": [
    {
      "id": "kubernetes",
      "label": "Kubernetes",
      "type": "technology",
      "layer": "UNIVERSAL",
      "attributes": {},
      "facts": [],
      "existing_id": null
    }
  ],
  "relations": [
    {"source": "beacon", "target": "kubernetes", "relation": "uses_technology"}
  ],
  "facts": [
    {"entity": "beacon", "text": "Has 50,000 users", "speaker": "user"}
  ]
}

Key capabilities

Capability	Description
Bilingual	Handles English and Spanish input natively
Empty output	Returns empty arrays for small talk and pure queries (no hallucinated entities)
Dedup awareness	References existing nodes via existing_id instead of creating duplicates
Implicit references	Maps "our project", "the app", "Alice's work" to existing graph nodes
Event extraction	Creates event nodes with participants, narrative position, and chronological markers
Controlled vocabulary	Uses strict enums for types (8) and relations (15)
Topic detection	Classifies same/subtopic/changed with optional hint from upstream classifiers

Training details

Parameter	Value
Base model	Qwen/Qwen3.5-9B
Method	LoRA (QLoRA 4-bit)
Framework	Unsloth + Transformers
Dataset size	~582 examples (450 base + 112 supplementary + 20 stress test)
Training	Initial 3 epochs (lr=2e-4) + incremental 2 epochs (lr=5e-5)
Max sequence length	2048
Languages	English (70%), Spanish (30%)

Dataset composition

Category	%	Description
Facts about existing entities	30%	"Our project has 50k users" → fact on existing node
New entity extraction	20%	First mentions of people, projects, technologies
Empty output (small talk / queries)	15%	"Thanks!", "What tech does X use?" → []
Topic changes	10%	Implicit and explicit topic switches
Subtopic shifts	10%	Diving deeper into an aspect
Literary events	5%	Events with narrative_position and chronological_marker
Corrections / updates	5%	"We switched from React to Vue"
Dedup / existing references	5%	"nuestro proyecto" → existing_id: "beacon"

Schema

Entity types (enum)

person, organization, project, technology, place, event, document, concept

Relation types (enum)

part_of, created_by, maintains, works_at, member_of,
uses_technology, depends_on, alternative_to,
located_in, deployed_on, produces, serves, documented_in,
participated_in, triggered_by, resulted_in

Layers

• PERSONAL — user owns, created, or directly uses it
• UNIVERSAL — public knowledge (technologies, fictional characters, cities)

Usage

With Transformers + LoRA

from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer

base_model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3.5-9B", device_map="auto")
model = PeftModel.from_pretrained(base_model, "sandyeveliz/acervo-extractor-qwen3.5-9b")
tokenizer = AutoTokenizer.from_pretrained("sandyeveliz/acervo-extractor-qwen3.5-9b")

messages = [
    {"role": "system", "content": "You are a knowledge extractor for a personal knowledge graph. Analyze the conversation and return a single JSON object with topic classification, entities, relations, and facts. Output valid JSON only, no markdown, no explanation."},
    {"role": "user", "content": "EXISTING NODES:\n[]\n\nTOPIC HINT: unresolved\nCURRENT TOPIC: null\n\nPREVIOUS ASSISTANT: null\nUSER: I work at Acme Corp building a React app called Beacon with PostgreSQL."}
]

inputs = tokenizer.apply_chat_template(messages, return_tensors="pt", add_generation_prompt=True)
outputs = model.generate(inputs.to(model.device), max_new_tokens=1024, temperature=0.1)
print(tokenizer.decode(outputs[0][inputs.shape[-1]:], skip_special_tokens=True))

With Unsloth (recommended for inference)

from unsloth import FastLanguageModel

model, tokenizer = FastLanguageModel.from_pretrained(
    "sandyeveliz/acervo-extractor-qwen3.5-9b",
    max_seq_length=2048, load_in_4bit=True,
)
FastLanguageModel.for_inference(model)

With Acervo (intended use)

from acervo import Acervo, OpenAIClient

llm = OpenAIClient(base_url="http://localhost:1234/v1", model="acervo-extractor")
memory = Acervo(llm=llm, owner="user")

Intended use

This model is designed as the extraction component inside Acervo, a semantic compression layer for AI agents. It replaces general-purpose LLM calls for topic detection and entity extraction with a specialized, faster model.

It can also be used standalone for:

• Building knowledge graphs from conversations
• Structured entity/relation extraction from text
• Topic detection in multi-turn dialogues

License

Apache 2.0 — same as the base model.