PicoOraClaw: Ultra-Efficient AI Assistant in Go + OCI GenAI + Oracle AI Database based on PicoClaw

$10 Hardware · 10MB RAM · 1s Boot · Oracle AI Vector Search · OCI GenAI (default) · Ollama · Any OpenAI-compatible API

PicoOraClaw is a fork of PicoClaw that adds Oracle AI Database as a backend for persistent storage and semantic vector search. By default it uses OCI Generative AI (xAI Grok 4) for inference, with Ollama for local open-weight models and any OpenAI-compatible API as alternatives. The agent remembers facts and recalls them by meaning using in-database ONNX embeddings (no external embedding API required).

_{Under 10MB RAM. Runs on $10 hardware.}

Architecture at a Glance

Full interactive presentation: Open picooraclaw-presentation.html in your browser for all 24 slides with animations and keyboard navigation.

Title	By the Numbers
25 Packages	11 Channels
Hardware Tools	Oracle Schema

Installation

One-command install. Clone, configure, and run in a single step:

curl -fsSL https://raw.githubusercontent.com/jasperan/picooraclaw/main/install.sh | bash

Advanced options

Override install location:

PROJECT_DIR=/opt/myapp curl -fsSL https://raw.githubusercontent.com/jasperan/picooraclaw/main/install.sh | bash

Or install manually:

git clone https://github.com/jasperan/picooraclaw.git
cd picooraclaw
# See below for setup instructions

🦾 Demonstration

🛠️ Standard AI Workflows

🧩 Full-Stack Engineering	🧠 Oracle AI Memory	🔎 Web Search & Learning

Develop · Deploy · Scale	Remember · Recall · Persist	Discover · Insights · Trends

⏰ Scheduled Tasks & Reminders

Set reminders, run recurring tasks, automate workflows. Scheduled jobs are stored persistently in Oracle AI Database with full ACID guarantees.

Quickstart (5 minutes)

Everything you need: Go 1.25+ and one of the LLM backends below. Optionally add Docker for Oracle AI Database 26ai Free.

Step 1: Build

git clone https://github.com/jasperan/picooraclaw.git
cd picooraclaw
make build
./build/picooraclaw onboard

Step 2: Pick your LLM backend

PicoOraClaw supports three ways to connect to an LLM. Pick one:

Option A: OCI Generative AI (default)

Uses Oracle Cloud's hosted models (xAI Grok 4, Meta Llama, Cohere) via a local proxy. No GPU needed.

Prerequisites: Python 3.11+, a configured ~/.oci/config profile.

# 1. Install the proxy
cd oci-genai && pip install -r requirements.txt && cd ..

# 2. Start the proxy (reads ~/.oci/config automatically)
export OCI_COMPARTMENT_ID=$(awk '/^tenancy=/{sub(/^tenancy=/,"");print;exit}' ~/.oci/config)
python3 oci-genai/proxy.py &

The default config already points at http://localhost:9999/v1 with xai.grok-4. No config edits needed.

Available models: xai.grok-4, xai.grok-3-mini, meta.llama-3.3-70b-instruct, cohere.command-r-plus. Check OCI GenAI docs for your region's availability.

Option B: Ollama (local open-weight models)

Runs models locally on your hardware. No cloud, no API keys, full privacy.

Prerequisites: Ollama installed, a GPU with enough VRAM for your chosen model.

# 1. Pull a model
ollama pull gemma4:26b    # 26B MoE, 17GB, needs 24GB+ VRAM
# or: ollama pull gemma4  # 8B, 9.6GB, fits on most GPUs

# 2. Edit ~/.picooraclaw/config.json

{
  "agents": {
    "defaults": {
      "provider": "ollama",
      "model": "gemma4:26b"
    }
  },
  "providers": {
    "ollama": {
      "api_key": "",
      "api_base": "http://localhost:11434/v1"
    }
  }
}

Option C: Any OpenAI-compatible API

Works with OpenAI, OpenRouter, Groq, Anthropic, DeepSeek, or any service that exposes a /v1/chat/completions endpoint.

{
  "agents": {
    "defaults": {
      "provider": "openai",
      "model": "gpt-4o"
    }
  },
  "providers": {
    "openai": {
      "api_key": "sk-your-key-here",
      "api_base": "https://api.openai.com/v1"
    }
  }
}

Swap api_base for any compatible endpoint (e.g., https://openrouter.ai/api/v1, https://api.groq.com/openai/v1). See the full provider list below.

Step 3: Chat

# One-shot
./build/picooraclaw agent -m "Hello!"

# Interactive mode
./build/picooraclaw agent

That's it. Use interactive mode when you want an ongoing local session.

Deploy to Oracle Cloud (One-Click)

Deploy a fully configured PicoOraClaw instance on OCI with Oracle AI Database, Ollama, and the gateway (all automated).

What gets deployed:

OCI Compute instance (shape of your choice, ARM A1.Flex is Always Free)
Ollama with qwen2.5:3b pre-loaded for CPU inference
Oracle AI Database 26ai Free container by default (or optional Autonomous AI Database when toggled)
PicoOraClaw gateway running as a systemd service

After deployment (~5-8 min for setup to complete):

# Check setup progress
ssh opc@<public_ip> -t 'tail -f /var/log/picooraclaw-setup.log'

# Start chatting
ssh opc@<public_ip> -t picooraclaw agent

# Check gateway health
curl http://<public_ip>:18790/health

Add Oracle AI Vector Search

Oracle gives you persistent storage, semantic memory (remember/recall by meaning), and crash-safe ACID transactions. Without it, storage is file-based.

Run the setup script. It handles everything automatically:

./scripts/setup-oracle.sh [optional-password]

This single script:

Pulls and starts the Oracle AI Database 26ai Free container
Waits for the database to be ready
Creates the picooraclaw database user with the required grants
Patches your ~/.picooraclaw/config.json with the Oracle connection settings
Downloads and stages the ONNX embedding model
Runs picooraclaw setup-oracle to initialize the schema and load the ONNX embedding model

Expected output when complete:

── Step 6/6: Schema + ONNX model ─────────────────────────────────────────
  Running picooraclaw setup-oracle...
✓ Connected to Oracle Database
✓ Schema initialized (8 tables with PICO_ prefix)
✓ ONNX model 'ALL_MINILM_L12_V2' already loaded
✓ VECTOR_EMBEDDING() test passed
✓ Prompts seeded from workspace

════════════════════════════════════════════════════════
  Oracle AI Database setup complete!
  Test with:
    ./build/picooraclaw agent -m "Remember that I love Go"
    ./build/picooraclaw agent -m "What language do I like?"
    ./build/picooraclaw oracle-inspect
════════════════════════════════════════════════════════

Step 5: Test semantic memory

# Store a fact
./build/picooraclaw agent -m "Remember that my favorite language is Go"

# Recall by meaning (not keywords)
./build/picooraclaw agent -m "What programming language do I prefer?"

The second command finds the stored memory via cosine similarity on 384-dimensional vectors (no keyword matching).

Step 6: Inspect what's stored

The oracle-inspect command lets you view everything stored in Oracle without writing SQL.

picooraclaw oracle-inspect [table] [options]

Tables: memories, sessions, transcripts, state, notes, prompts, config, meta Options: -n <limit> max rows (default 20), -s <text> semantic search (memories only)

Overview dashboard (no arguments)

./build/picooraclaw oracle-inspect

=============================================================
  PicoOraClaw Oracle AI Database Inspector
=============================================================

  Table                  Rows
  ─────────────────────  ────
  Memories                  20  ████████████████████
  Sessions                   4  ████
  Transcripts                6  ██████
  State                      8  ████████
  Daily Notes                3  ███
  Prompts                    4  ████
  Config                     2  ██
  Meta                       1  █
  ─────────────────────  ────
  Total                     48

  Recent Memories (last 5):
  ─────────────────────────────────────────────────────────
  2026-02-19 04:13  0.6 [preference]  Design docs and plan files go in docs/plans/ ...
  2026-02-19 04:13  0.6 [interest]  User is interested in IoT and embedded systems ...
  2026-02-19 04:13  0.7 [fact]  The Oracle ONNX embedding model used is ALL_MINILM_L12_V2 ...
  2026-02-19 04:13  0.7 [preference]  For complex multi-step tasks, use multi-agent parallel ...
  2026-02-19 04:13  0.8 [preference]  User prefers concise communication ...

  Recent Transcripts (last 5):
  ─────────────────────────────────────────────────────────
  2026-02-19 04:13  assistant   [discord:dev-channel]  I've stored that in memory ...
  2026-02-19 04:13  user        [discord:dev-channel]  Remember that the next release ...
  2026-02-19 04:13  assistant   [telegram:user123]  Oracle Database connection is active ...
  2026-02-19 04:13  user        [telegram:user123]  What's the status of the Oracle connection?
  2026-02-19 04:13  assistant   [cli:repl-session]  I can help with that ...

  Recent Sessions (last 5):
  ─────────────────────────────────────────────────────────
  2026-02-19 04:13  discord:dev-channel             (no summary)
  2026-02-19 04:13  telegram:user123                (no summary)
  2026-02-19 04:13  cli:repl-session                (no summary)
  2026-02-18 06:07  cli:default                     **Cohesive Summary:** The user and
  assistant engaged in a playful exchange starting with "Pong!" ...

  Recent State Entries (last 5):
  ─────────────────────────────────────────────────────────
  2026-02-19 04:13  total_conversations       = 42
  2026-02-19 04:13  agent_mode                = interactive
  2026-02-19 04:13  user_timezone             = America/Los_Angeles
  2026-02-19 04:13  user_name                 = jasperan
  2026-02-19 04:13  last_chat_id              = repl-session

  Recent Daily Notes (last 5):
  ─────────────────────────────────────────────────────────
  2026-02-19  (updated 2026-02-19 04:13)  # 2026-02-19 ...
  2026-02-18  (updated 2026-02-19 04:13)  # 2026-02-18 ...
  2026-02-17  (updated 2026-02-19 04:13)  # 2026-02-17 ...

  System Prompts (last 5):
  ─────────────────────────────────────────────────────────
  2026-02-18 08:38  AGENT                      357 chars
  2026-02-18 08:38  USER                       365 chars
  2026-02-18 08:38  SOUL                       296 chars
  2026-02-18 08:38  IDENTITY                  1271 chars

  Config Entries (last 5):
  ─────────────────────────────────────────────────────────
  2026-02-19 04:13  last_seed                 = 2026-02-19T04:13:34Z
  2026-02-19 04:13  full_config               = { "llm": {...}, "oracle": {...}, ... }

  Schema Metadata:
  ─────────────────────────────────────────────────────────
  2026-02-18 08:38  schema_version                 = 1.0.0

  Tip: Run 'picooraclaw oracle-inspect <table>' for details
       Run 'picooraclaw oracle-inspect memories -s "query"' for semantic search

List all memories

./build/picooraclaw oracle-inspect memories

  All Memories
  ─────────────────────────────────────────────────────────

  ID: faffd019  Vector: yes
  Created: 2026-02-19 04:13  Importance: 0.9  Category: preference  Accessed: 0x
  Content: User prefers Oracle Database as the primary database. They work at Oracle
  and prefer Oracle AI Vector Search for embeddings.

  ID: 0e39036f  Vector: yes
  Created: 2026-02-19 04:13  Importance: 0.8  Category: preference  Accessed: 0x
  Content: Go is the user's primary programming language. They use Go 1.24 and target
  embedded Linux devices (RISC-V, ARM64, x86_64).

  ID: 7aca4a7b  Vector: yes
  Created: 2026-02-19 04:13  Importance: 0.8  Category: preference  Accessed: 0x
  Content: User prefers Ollama as the open-source LLM framework for local inference.

  ID: 27c2473c  Vector: yes
  Created: 2026-02-19 04:13  Importance: 0.7  Category: interest  Accessed: 0x
  Content: User is interested in RAG (Retrieval-Augmented Generation) using LlamaIndex
  and LangChain frameworks.

  ...

Semantic search over memories

./build/picooraclaw oracle-inspect memories -s "what does the user like to program in"

  Semantic Search: "what does the user like to program in"
  ─────────────────────────────────────────────────────────

  [ 61.3% match]  ID: 383ff5d3
  Created: 2026-02-16 06:13  Importance: 0.7  Category: preference  Accessed: 0x
  Content: I prefer Python and Go for programming

  [ 60.7% match]  ID: 0e74a94c
  Created: 2026-02-18 02:20  Importance: 0.7  Category: preference  Accessed: 0x
  Content: my favorite programming language is Go

  [ 40.1% match]  ID: 0e39036f
  Created: 2026-02-19 04:13  Importance: 0.8  Category: preference  Accessed: 0x
  Content: Go is the user's primary programming language. They use Go 1.24 and target
  embedded Linux devices (RISC-V, ARM64, x86_64).

  [ 30.9% match]  ID: 22b84dba
  Created: 2026-02-16 06:12  Importance: 0.7  Category: employment  Accessed: 1x
  Content: I work at Oracle as a developer

  ...

Inspect sessions

./build/picooraclaw oracle-inspect sessions

  Chat Sessions
  ─────────────────────────────────────────────────────────

  Session: discord:dev-channel
  Created: 2026-02-19 04:13  Updated: 2026-02-19 04:13  Messages size: 673 bytes

  Session: telegram:user123
  Created: 2026-02-19 04:13  Updated: 2026-02-19 04:13  Messages size: 304 bytes

  Session: cli:repl-session
  Created: 2026-02-19 04:13  Updated: 2026-02-19 04:13  Messages size: 982 bytes

  Session: cli:default
  Created: 2026-02-16 06:12  Updated: 2026-02-18 06:07  Messages size: 2848 bytes
  Summary: **Cohesive Summary:** The user and assistant engaged in a playful
  exchange starting with "Pong!" The assistant recalled the user's role as an
  Oracle developer and their preference for Go ...

Inspect agent state

./build/picooraclaw oracle-inspect state

  Agent State (Key-Value)
  ─────────────────────────────────────────────────────────
  agent_mode                     = interactive                    (2026-02-19 04:13)
  last_channel                   = cli                            (2026-02-19 04:13)
  last_chat_id                   = repl-session                   (2026-02-19 04:13)
  last_model                     = gpt-4o-mini                    (2026-02-19 04:13)
  tools_used_count               = 187                            (2026-02-19 04:13)
  total_conversations            = 42                             (2026-02-19 04:13)
  user_name                      = jasperan                       (2026-02-19 04:13)
  user_timezone                  = America/Los_Angeles            (2026-02-19 04:13)

Inspect daily notes

./build/picooraclaw oracle-inspect notes

  Daily Notes
  ─────────────────────────────────────────────────────────

  Date: 2026-02-19  ID: 2ccb7e70  Vector: yes  Updated: 2026-02-19 04:13
  Content: # 2026-02-19
  ## Development Progress
  - Implemented seed-demo command for Oracle data population
  - Tested vector embeddings with ALL_MINILM_L12_V2 model
  - Fixed session serialization edge case with empty tool calls

  Date: 2026-02-18  ID: 29a60563  Vector: yes  Updated: 2026-02-19 04:13
  Content: # 2026-02-18
  ## Testing & Debugging
  - Ran full test suite: all 47 tests passing
  - Profiled memory usage on RISC-V board: 8.2MB peak
  - Verified Oracle connection pooling under concurrent load

  Date: 2026-02-17  ID: f4cfb628  Vector: no  Updated: 2026-02-19 04:13
  Content: # 2026-02-17
  ## Architecture Planning
  - Designed transcript storage schema for PICO_TRANSCRIPTS table
  - Sketched WhatsApp channel adapter following existing Telegram pattern
  - Reviewed LangChain-OracleDB integration for RAG pipeline

Inspect transcripts

./build/picooraclaw oracle-inspect transcripts

  Conversation Transcripts
  ─────────────────────────────────────────────────────────
  2026-02-19 04:13  #2  assistant  [discord:dev-channel]   I've stored that in memory ...
  2026-02-19 04:13  #1  user       [discord:dev-channel]   Remember that the next release ...
  2026-02-19 04:13  #2  assistant  [telegram:user123]      Oracle Database connection is active ...
  2026-02-19 04:13  #1  user       [telegram:user123]      What's the status of the Oracle connection?
  2026-02-19 04:13  #2  assistant  [cli:repl-session]      I can help with that ...
  2026-02-19 04:13  #1  user       [cli:repl-session]      Can you help me add a new tool ...

Inspect stored config

./build/picooraclaw oracle-inspect config

  Stored Config
  ─────────────────────────────────────────────────────────
  full_config                    = {
  "llm": {"provider": "openai-compatible", "model": "gpt-4o-mini", ...},
  "oracle": {"enabled": true, "onnx_model": "ALL_MINILM_L12_V2"},
  "channels": {"telegram": {"enabled": true}, "discord": {"enabled": true}},
  "agent": {"max_tool_iterations": 10, "context_window": 8192}
}  (2026-02-19 04:13)
  last_seed                      = 2026-02-19T04:13:34Z       (2026-02-19 04:13)

View a system prompt in full

./build/picooraclaw oracle-inspect prompts IDENTITY
./build/picooraclaw oracle-inspect prompts SOUL

Schema metadata, ONNX models, and vector indexes

./build/picooraclaw oracle-inspect meta

  Schema Metadata
  ─────────────────────────────────────────────────────────
  schema_version                 = 1.0.0

  ONNX Models
  ─────────────────────────────────────────────────────────
  ALL_MINILM_L12_V2          EMBEDDING        ONNX

  Vector Indexes
  ─────────────────────────────────────────────────────────
  IDX_PICO_DAILY_NOTES_VEC        on PICO_DAILY_NOTES
  IDX_PICO_MEMORIES_VEC           on PICO_MEMORIES

CLI Reference

Command	Description
`picooraclaw onboard`	Initialize config and workspace
`picooraclaw agent -m "..."`	One-shot chat
`picooraclaw agent`	Interactive chat mode
`picooraclaw gateway`	Start long-running service with channels
`picooraclaw status`	Show status
`picooraclaw setup-oracle`	Initialize Oracle schema + ONNX model
`picooraclaw oracle-inspect`	Inspect data stored in Oracle
`picooraclaw oracle-inspect memories -s "query"`	Semantic search over memories
`picooraclaw seed-demo`	Populate Oracle with realistic demo data
`picooraclaw cron list`	List scheduled jobs
`picooraclaw skills list`	List installed skills

How Oracle Storage Works

                           ┌──────────────────────────────────────────┐
                           │         Oracle AI Database               │
                           │                                          │
  picooraclaw binary       │  ┌──────────────┐  ┌──────────────────┐ │
  ┌───────────────────┐    │  │ PICO_MEMORIES │  │ PICO_DAILY_NOTES │ │
  │  AgentLoop        │    │  │  + VECTOR idx │  │  + VECTOR idx    │ │
  │  ├─ SessionStore ──────│──│──────────────┐│  └──────────────────┘ │
  │  ├─ StateStore   ──────│──│ PICO_SESSIONS││                       │
  │  ├─ MemoryStore  ──────│──│ PICO_STATE   ││  ┌──────────────────┐ │
  │  ├─ PromptStore  ──────│──│ PICO_PROMPTS ││  │ ALL_MINILM_L12_V2│ │
  │  ├─ ConfigStore  ──────│──│ PICO_CONFIG  ││  │   (ONNX model)   │ │
  │  └─ Tools:       │    │  │ PICO_META    ││  │  384-dim vectors  │ │
  │     ├─ remember  ──────│──│ PICO_TRANS.  ││  └──────────────────┘ │
  │     └─ recall    ──────│──└──────────────┘│                       │
  └───────────────────┘    │   go-ora v2.9.0  │                       │
         (pure Go)         │   (pure Go driver)│                       │
                           └──────────────────────────────────────────┘

Table	Purpose
`PICO_MEMORIES`	Long-term memory with 384-dim vector embeddings for semantic search
`PICO_SESSIONS`	Chat history per channel
`PICO_TRANSCRIPTS`	Full conversation audit log
`PICO_STATE`	Agent key-value state
`PICO_DAILY_NOTES`	Daily journal entries with vector embeddings
`PICO_PROMPTS`	System prompts (IDENTITY.md, SOUL.md, etc.)
`PICO_CONFIG`	Runtime configuration
`PICO_META`	Schema versioning metadata

The remember tool stores text + vector embedding via VECTOR_EMBEDDING(ALL_MINILM_L12_V2 USING :text AS DATA). The recall tool searches by cosine similarity via VECTOR_DISTANCE(). Results with < 30% similarity are filtered out.

Using Alternative LLM Providers

PicoOraClaw defaults to OCI GenAI, but you can swap to any provider by editing ~/.picooraclaw/config.json. Set provider and add your API key:

OpenRouter (access to all models)

{
  "agents": {
    "defaults": {
      "provider": "openrouter",
      "model": "anthropic/claude-opus-4-5"
    }
  },
  "providers": {
    "openrouter": {
      "api_key": "sk-or-v1-xxx",  # pragma: allowlist secret
      "api_base": "https://openrouter.ai/api/v1"
    }
  }
}

Get a key at openrouter.ai/keys (200K free tokens/month).

All supported providers

Provider	Purpose	Get API Key
`openai` + OCI proxy	OCI GenAI (default) (xAI Grok, Llama, Cohere)	OCI credentials
`ollama`	Local open-weight inference (no cloud)	ollama.com
`openai`	GPT models / any OpenAI-compatible API	platform.openai.com
`openrouter`	Access to all models	openrouter.ai
`anthropic`	Claude models	console.anthropic.com
`gemini`	Gemini models	aistudio.google.com
`groq`	Fast inference + voice transcription	console.groq.com
`deepseek`	DeepSeek models	platform.deepseek.com
`nvidia`	NVIDIA-hosted OpenAI-compatible models	build.nvidia.com
`vllm`	Self-hosted vLLM-compatible endpoint	Set your own API base
`moonshot`	Moonshot/Kimi models via model-name detection	platform.moonshot.cn
`github_copilot`	GitHub Copilot provider	GitHub Copilot entitlement

OCI Generative AI (Default Backend)

PicoOraClaw uses OCI Generative AI as its default LLM backend. A lightweight Python proxy (oci-genai/proxy.py) translates standard OpenAI API calls into OCI-authenticated requests, so the Go binary stays dependency-free.

PicoOraClaw (Go) --> localhost:9999/v1 (proxy.py) --> OCI GenAI endpoint
                     OpenAI-compatible              OCI User Principal Auth

Available OCI GenAI Models

Model ID	Description
`xai.grok-4`	xAI Grok 4 (default, strong tool calling)
`xai.grok-3-mini`	xAI Grok 3 Mini (fast, detailed reasoning traces)
`meta.llama-3.3-70b-instruct`	Meta Llama 3.3 70B Instruct
`cohere.command-r-plus`	Cohere Command R+
`meta.llama-3.1-405b-instruct`	Meta Llama 3.1 405B Instruct

Model availability varies by region. Check the OCI GenAI documentation for the latest list.

Setup

Install the proxy:

cd oci-genai && pip install -r requirements.txt

Configure OCI credentials (~/.oci/config):

[DEFAULT]
user=ocid1.user.oc1..aaaaaaaaexample
fingerprint=aa:bb:cc:dd:ee:ff:00:11:22:33:44:55:66:77:88:99
tenancy=ocid1.tenancy.oc1..aaaaaaaaexample
region=us-chicago-1
key_file=~/.oci/oci_api_key.pem

Start the proxy (auto-reads ~/.oci/config):

export OCI_COMPARTMENT_ID=$(awk '/^tenancy=/{sub(/^tenancy=/,"");print;exit}' ~/.oci/config)
python3 oci-genai/proxy.py
# Proxy runs at http://localhost:9999/v1

The default config already points to this endpoint. No config edits needed.

Run the OCI demo (starts proxy, runs 5 beats, cleans up):

./demo-oci.sh                   # xai.grok-4 (default)
./demo-oci.sh xai.grok-3-mini   # faster, with reasoning traces

See oci-genai/README.md for full documentation.

Chat Channels

Connect PicoOraClaw to Telegram, WhatsApp, Feishu, Discord, MaixCam, QQ, DingTalk, Slack, LINE, OneBot, or Web via the gateway command.

Telegram (Recommended)

Message @BotFather on Telegram, send /newbot, copy the token
Add to ~/.picooraclaw/config.json:

{
  "channels": {
    "telegram": {
      "enabled": true,
      "token": "YOUR_BOT_TOKEN",
      "allow_from": ["YOUR_USER_ID"]
    }
  }
}

Run picooraclaw gateway

Discord

Create a bot at discord.com/developers, enable MESSAGE CONTENT INTENT
Add to config:

{
  "channels": {
    "discord": {
      "enabled": true,
      "token": "YOUR_BOT_TOKEN",
      "allow_from": ["YOUR_USER_ID"]
    }
  }
}

Invite bot with Send Messages + Read Message History permissions
Run picooraclaw gateway

WhatsApp, Feishu, MaixCam, QQ, DingTalk, Slack, LINE, OneBot, Web

See config/config.example.json for the full channel configuration reference. Each channel follows the same pattern:

{
  "channels": {
    "<channel_name>": {
      "enabled": true,
      "<credentials>": "...",
      "allow_from": []
    }
  }
}

Run picooraclaw gateway after configuring.

Oracle on Autonomous AI Database (Cloud, Optional)

ADB wallet-less TLS

{
  "oracle": {
    "enabled": true,
    "mode": "adb",
    "dsn": "(description=(retry_count=20)(retry_delay=3)(address=(protocol=tcps)(port=1522)(host=adb.us-ashburn-1.oraclecloud.com))(connect_data=(service_name=xxx_myatp_low.adb.oraclecloud.com))(security=(ssl_server_dn_match=yes)))",
    "user": "picooraclaw",
    "password": "YourPass123"  # pragma: allowlist secret
  }
}

ADB with mTLS wallet

{
  "oracle": {
    "enabled": true,
    "mode": "adb",
    "host": "adb.us-ashburn-1.oraclecloud.com",
    "port": 1522,
    "service": "xxx_myatp_low.adb.oraclecloud.com",
    "wallet_path": "/path/to/wallet",
    "user": "picooraclaw",
    "password": "YourPass123"  # pragma: allowlist secret
  }
}

Download wallet from OCI Console > Autonomous Database > DB Connection > Download Wallet.

Oracle config reference

Field	Env Variable	Default	Description
`enabled`	`PICO_ORACLE_ENABLED`	`false`	Enable Oracle backend
`mode`	`PICO_ORACLE_MODE`	`freepdb`	`freepdb` or `adb`
`host`	`PICO_ORACLE_HOST`	`localhost`	Oracle host
`port`	`PICO_ORACLE_PORT`	`1521`	Listener port
`service`	`PICO_ORACLE_SERVICE`	`FREEPDB1`	Service name
`user`	`PICO_ORACLE_USER`	`picooraclaw`	DB username
`password`	`PICO_ORACLE_PASSWORD`	(none)	DB password
`dsn`	`PICO_ORACLE_DSN`	(none)	Full DSN (ADB wallet-less)
`wallet_path`	`PICO_ORACLE_WALLET_PATH`	(none)	Wallet directory (ADB mTLS)
`pool_max_open`	`PICO_ORACLE_POOL_MAX_OPEN`	`10`	Maximum open DB connections
`pool_max_idle`	`PICO_ORACLE_POOL_MAX_IDLE`	`2`	Maximum idle DB connections
`onnx_model`	`PICO_ORACLE_ONNX_MODEL`	`ALL_MINILM_L12_V2`	ONNX model for embeddings
`agent_id`	`PICO_ORACLE_AGENT_ID`	`default`	Multi-agent isolation key
`embedding_provider`	`PICO_ORACLE_EMBEDDING_PROVIDER`	`api`	`api` with an API key, otherwise ONNX fallback
`embedding_api_base`	`PICO_ORACLE_EMBEDDING_API_BASE`	(none)	OpenAI-compatible embeddings API base
`embedding_api_key`	`PICO_ORACLE_EMBEDDING_API_KEY`	(none)	Embeddings API key
`embedding_model`	`PICO_ORACLE_EMBEDDING_MODEL`	`embedding-3`	Embeddings API model

Troubleshooting

Oracle: Connection refused / ORA-12541

docker ps | grep oracle          # Is it running?
docker logs oracle-free          # Wait for "DATABASE IS READY"
ss -tlnp | grep 1521            # Is port 1521 listening?

Oracle: ORA-01017 invalid username/password

docker exec -it oracle-free sqlplus sys/YourPass123@localhost:1521/FREEPDB1 as sysdba
SQL> ALTER USER picooraclaw IDENTIFIED BY NewPassword123;

Oracle: VECTOR_EMBEDDING() returns ORA-04063

ONNX model not loaded. Run picooraclaw setup-oracle or manually:

BEGIN
  DBMS_VECTOR.LOAD_ONNX_MODEL('PICO_ONNX_DIR', 'all_MiniLM_L12_v2.onnx', 'ALL_MINILM_L12_V2');
END;
/

Requires GRANT CREATE MINING MODEL TO picooraclaw; as SYSDBA.

Agent falls back to file-based mode

Oracle is enabled but connection failed at startup. Check:

Is the Oracle container healthy? (docker ps)
Password match between config and PICO_ORACLE_PASSWORD or the setup script argument?
Service name should be FREEPDB1 (not FREE or XE)

Build Targets

make build          # Build for current platform
make build-all      # Cross-compile: linux/{amd64,arm64,riscv64}, darwin/arm64, windows/amd64
make install        # Build + install to ~/.local/bin
make test           # go test ./...
make fmt            # go fmt ./...
make vet            # go vet ./...

Docker Compose

# Full stack with Oracle
PICO_ORACLE_PASSWORD=YourPass123 docker compose --profile oracle --profile gateway up -d  # pragma: allowlist secret

# Without Oracle
docker compose --profile gateway up -d

# One-shot agent
docker compose run --rm picooraclaw-agent -m "What is 2+2?"

Features

Single static binary (~10MB RAM), runs on RISC-V/ARM64/x86_64
Ollama, OpenRouter, Anthropic, OpenAI, Gemini, DeepSeek, Groq, NVIDIA, vLLM, Moonshot, and GitHub Copilot providers
Optional Oracle AI Database 26ai Free backend with AI Vector Search (384-dim ONNX embeddings)
Chat channels: Telegram, WhatsApp, Feishu, Discord, MaixCam, QQ, DingTalk, Slack, LINE, OneBot, Web
Scheduled tasks via cron expressions
Heartbeat periodic tasks
Skills system (workspace, global, GitHub-hosted)
Security sandbox with workspace restriction
Optional: Oracle Autonomous AI Database for managed cloud deployment
Graceful fallback to file-based storage when Oracle is unavailable

Name		Name	Last commit message	Last commit date
Latest commit History 82 Commits
.omc		.omc
.serena		.serena
assets		assets
cmd/picooraclaw		cmd/picooraclaw
config		config
deploy/oci		deploy/oci
docs/slides		docs/slides
oci-genai		oci-genai
pkg		pkg
remotion		remotion
scripts		scripts
workspace		workspace
.dockerignore		.dockerignore
.gitignore		.gitignore
.goreleaser.yaml		.goreleaser.yaml
.pre-commit-config.yaml		.pre-commit-config.yaml
.python-version		.python-version
.secrets.baseline		.secrets.baseline
Dockerfile		Dockerfile
Dockerfile.goreleaser		Dockerfile.goreleaser
LICENSE		LICENSE
Makefile		Makefile
README.ja.md		README.ja.md
README.md		README.md
README.zh.md		README.zh.md
demo-oci.sh		demo-oci.sh
demo.sh		demo.sh
docker-compose.yml		docker-compose.yml
go.mod		go.mod
go.sum		go.sum
install.sh		install.sh
picooraclaw-presentation.html		picooraclaw-presentation.html
pyproject.toml		pyproject.toml
uv.lock		uv.lock

Folders and files

Latest commit

History

Repository files navigation

PicoOraClaw: Ultra-Efficient AI Assistant in Go + OCI GenAI + Oracle AI Database based on PicoClaw

$10 Hardware · 10MB RAM · 1s Boot · Oracle AI Vector Search · OCI GenAI (default) · Ollama · Any OpenAI-compatible API

Architecture at a Glance

Installation

🦾 Demonstration

🛠️ Standard AI Workflows

⏰ Scheduled Tasks & Reminders

Quickstart (5 minutes)

Step 1: Build

Step 2: Pick your LLM backend

Step 3: Chat

Deploy to Oracle Cloud (One-Click)

Add Oracle AI Vector Search

Step 5: Test semantic memory

Step 6: Inspect what's stored

Overview dashboard (no arguments)

List all memories

Semantic search over memories

Inspect sessions

Inspect agent state

Inspect daily notes

Inspect transcripts

Inspect stored config

View a system prompt in full

Schema metadata, ONNX models, and vector indexes

CLI Reference

How Oracle Storage Works

Using Alternative LLM Providers

OCI Generative AI (Default Backend)

Available OCI GenAI Models

Setup

Chat Channels

Oracle on Autonomous AI Database (Cloud, Optional)

Troubleshooting

Build Targets

Docker Compose

Features

About

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages