Skip to content

basic-intent.md

Latest commit

 

History

History
105 lines (73 loc) · 3.25 KB

basic-intent.md

File metadata and controls

105 lines (73 loc) · 3.25 KB

BasicIntentAnalyzer — Intent analysis (with deception indicators)

BasicIntentAnalyzer is a structured intent-analysis component built on AbstractCore. It can analyze:

  • a single text input (analyze_intent)
  • a multi-participant conversation (analyze_conversation_intents)

Deception indicators are always included in the structured output schema.

Install

# Core (lightweight)
pip install abstractcore

# Provider extras (install only what you use)
pip install "abstractcore[openai]"
pip install "abstractcore[anthropic]"
pip install "abstractcore[huggingface]"  # heavy (torch/transformers)

# Optional: read PDFs / Office docs in the CLI
pip install "abstractcore[media]"

Quick start (Python)

from abstractcore import create_llm
from abstractcore.processing import BasicIntentAnalyzer, IntentContext, IntentDepth

llm = create_llm("openai", model="gpt-4o-mini")  # requires `abstractcore[openai]`
analyzer = BasicIntentAnalyzer(llm)

result = analyzer.analyze_intent(
    "I'm struggling to understand this concept and need help.",
    context_type=IntentContext.STANDALONE,
    depth=IntentDepth.UNDERLYING,
)

print(result.primary_intent.intent_type.value)
print(result.primary_intent.underlying_goal)
print(result.primary_intent.deception_analysis.deception_likelihood)

CLI usage (intent)

The intent CLI analyzes either direct text or a file path:

# Direct text
intent "I need help with this problem" --depth underlying

# File input
intent ./email.txt --context document --depth comprehensive --format plain

# Conversation log input (expects lines like `USER: ...` / `ASSISTANT: ...`)
intent ./chat.txt --conversation-mode --focus-participant user --format plain

# Session archive JSON created via `BasicSession.save(...)` (auto-enables conversation mode)
intent ./saved_session.json --focus-participant user --depth comprehensive --format plain

Key flags

  • --context: standalone, conversational, document, interactive
  • --depth: surface, underlying, comprehensive
  • --conversation-mode: parse multi-message inputs; use --focus-participant to target a role
  • --provider + --model: override the default app model (must be provided together)
  • --format: json, yaml, plain
  • --timeout: HTTP timeout in seconds (0 = unlimited)

Tip: to see every option, run intent --help.

Interactive chat integration (/intent)

The interactive chat CLI (abstractcore-chat) supports /intent to analyze the current conversation state:

abstractcore-chat --provider ollama --model gemma3:1b-it-qat

# In the REPL:
/intent
/intent user
/intent assistant

Events (optional)

You can monitor generations via the global event bus:

from abstractcore.events import EventType, on_global

def monitor(event):
    if event.type == EventType.GENERATION_COMPLETED:
        duration_ms = event.data.get("duration_ms")
        print("generation_completed", duration_ms)

on_global(EventType.GENERATION_COMPLETED, monitor)

Notes

  • Long inputs are chunked and combined (map-reduce). Use --chunk-size in the CLI (or max_chunk_size in Python) to control chunk size.
  • Output schemas live in abstractcore/processing/basic_intent.py (IntentAnalysisOutput, IntentType, DeceptionIndicators).