🧮 CLI Commands

Prefer terminals over tabs? Locentra OS ships with a full CLI interface that lets you query, train, vectorize, and manage without ever opening a browser or hitting an API endpoint.

All CLI scripts live in:

backend/cli/

Run them like this:

python cli/<command>.py [arguments]

Each script has access to:

• 🔁 Locentra Registry

• 🧠 Vector memory engine

• 🔧 Config settings (.env)

• 🧾 Logger system

• 🧱 Model + tokenizer

---

📥 query.py

Send a prompt to the active model and print the output to your terminal.

Usage:

python cli/query.py --prompt "Summarize LSDfi"

Options:

Flag	Description
--prompt	The text prompt to send to the model
--max_tokens	Maximum number of tokens to generate
--temperature	Sampling temperature (creativity control)
--verbose	Show internal debug info and memory scores

Output:

[Model Output]
LSDfi refers to DeFi protocols built around Liquid Staking Derivatives...

---

🧠 train.py

Push new knowledge into the model using prompt → completion training.

Basic Example:

python cli/train.py \
  --prompt "What is TVL?" \
  --completion "TVL stands for Total Value Locked..."

Flags:

Flag	Description
--prompt	Input prompt to train on
--completion	Expected response (used for supervised tuning)
--vectorize	Store the prompt in vector memory
--tags	Attach semantic tags for future filtering
--dry-run	Simulate training without applying it
--meta	Attach metadata (e.g. source, user, origin)

🧪 Tip: Automate batch training with:

bash scripts/train_batch.sh

---

🧹 clean_memory.py (Planned Module)

Optional memory housekeeping utility (suggested extension):

python cli/clean_memory.py --older-than 30d --tags "demo,test"

Concept:

• Delete or archive outdated memory entries

• Filter by vector score, timestamp, or tag

---

🧩 Integration Use Case:

Build your own CLI pipelines by chaining commands and external data sources.

🔹 Train from a scraped article:

curl https://some-article.ai | \
  grep "LLM" | \
  python cli/train.py --prompt "$(cat -)" --completion "..."

🔹 Log every query to disk:

python cli/query.py --prompt "Explain zkEVM" >> history.log

Locentra CLI is scriptable and Unix-friendly. Perfect for automated fine-tuning loops, daily prompt logging, or memory introspection.

---

🧰 Extend the CLI

Every script in backend/cli/ has access to:

• registry.get() and .register()

• memory_service.log_prompt()

• embed_text() for semantic recall

• fine_tune_model() for instant training

• All system settings via from backend.core.config import settings

Add your own:

backend/cli/my_custom_script.py

You’ll immediately inherit full access to the runtime system.