Furhat memory

.devcontainer/Dockerfile

@@ -1,55 +1,64 @@
-# Use an official Python 3.11 runtime as a base image
-FROM python:3.11-slim
+# -------------------------------------------------------------
+# Stage 1: Base Python environment with build tools
+# -------------------------------------------------------------
+    FROM python:3.11-slim
 
-# Install system dependencies, including build tools, git, cmake, clang, libc++-dev, libc++abi-dev, libomp-dev, ninja-build, Python development headers, OpenBLAS, and pkg-config
-RUN apt-get update && apt-get install -y --no-install-recommends \
-    build-essential \
-    git \
-    cmake \
-    clang \
-    libc++-dev \
-    libc++abi-dev \
-    libomp-dev \
-    ninja-build \
-    python3-dev \
-    libopenblas-dev \
-    pkg-config \
-  && rm -rf /var/lib/apt/lists/*
-
-# Upgrade pip, setuptools, and wheel, then install Poetry 2.0.1
-RUN pip install --upgrade pip setuptools wheel && \
-    pip install poetry==2.0.1
-
-# Set the working directory in the container
-WORKDIR /app
-
-# Copy dependency files first to leverage Docker cache
-COPY pyproject.toml poetry.lock* /app/
-
-# Install dependencies using Poetry without installing the root package
-RUN poetry config virtualenvs.create false && \
-    poetry install --no-root --no-interaction --no-ansi
-
-# (Optional) Copy and install additional dependencies from requirements_poetry.txt if present
-COPY requirements_poetry.txt /app/
-RUN if [ -f requirements_poetry.txt ]; then pip install --no-cache-dir -r requirements_poetry.txt; fi
-
-# # Set CMake arguments for OpenBLAS support
-# ENV CMAKE_ARGS="-DGGML_BLAS=ON -DGGML_BLAS_VENDOR=OpenBLAS"
-
-# # Install llama_cpp_python with verbose output
-# RUN pip install --no-cache-dir --verbose llama_cpp_python==0.3.7
-
-# RUN CMAKE_ARGS="-DGGML_BLAS=ON -DGGML_BLAS_VENDOR=OpenBLAS" pip install llama-cpp-python
-
-RUN CMAKE_ARGS="-DGGML_NATIVE=OFF -DGGML_CPU_ARM_ARCH=armv8-a" pip install llama-cpp-python
-
-
-# Copy the rest of the project files
-COPY . /app
-
-# Expose the port if needed (or you can omit if not running the server automatically)
-EXPOSE 8000
-
-# Instead of running the app automatically, start a shell for interactive work
-CMD [ "bash" ]
+    # System dependencies for building and for Ollama
+    RUN apt-get update && apt-get install -y --no-install-recommends \
+        build-essential git curl wget cmake clang libc++-dev libc++abi-dev \
+        libomp-dev ninja-build python3-dev libopenblas-dev pkg-config ca-certificates \
+      && rm -rf /var/lib/apt/lists/*
+
+    # -------------------------------------------------------------
+    # Install Poetry and project dependencies
+    # -------------------------------------------------------------
+    RUN pip install --upgrade pip setuptools wheel && pip install poetry==2.0.1
+    WORKDIR /app
+
+    # Copy Poetry files
+    COPY pyproject.toml poetry.lock* /app/
+    RUN poetry config virtualenvs.create false && poetry install --no-root --no-interaction --no-ansi
+
+    # Optional extra dependencies
+    COPY requirements_poetry.txt /app/
+    RUN if [ -f requirements_poetry.txt ]; then pip install --no-cache-dir -r requirements_poetry.txt; fi
+
+    # -------------------------------------------------------------
+    # Install Ollama
+    # -------------------------------------------------------------
+    # Ollama official Linux install script:
+    RUN curl -fsSL https://ollama.com/install.sh | sh
+
+    # Add Ollama binary to PATH (just to be sure)
+    ENV PATH="/usr/local/bin:${PATH}"
+
+    # -------------------------------------------------------------
+    # (Optional) Install llama-cpp-python if you still want local GGUF support
+    # Comment out if you’ll use Ollama exclusively
+    # RUN CMAKE_ARGS="-DGGML_BLAS=ON -DGGML_BLAS_VENDOR=OpenBLAS" pip install llama-cpp-python==0.3.7
+
+    # -------------------------------------------------------------
+    # Copy the rest of your app
+    # -------------------------------------------------------------
+    COPY . /app
+
+    # Environment variables
+    ENV OLLAMA_BASE_URL="http://localhost:11434"
+    ENV HF_HOME=/app/.cache/hf
+    ENV TRANSFORMERS_CACHE=/app/.cache/hf
+    ENV SENTENCE_TRANSFORMERS_HOME=/app/.cache/st
+
+    # Expose the FastAPI port
+    EXPOSE 8000
+
+    # -------------------------------------------------------------
+    # Entrypoint: start Ollama service + your API
+    # -------------------------------------------------------------
+    # Ollama runs as a background service; we then start FastAPI
+    CMD bash -c "\
+        echo 'Starting Ollama service...' && \
+        ollama serve & \
+        sleep 3 && \
+        echo 'Starting FastAPI app...' && \
+        uvicorn main:app --host 0.0.0.0 --port 8000 \
+    "

README.md

@@ -261,6 +261,18 @@ The project includes several GPU optimizations:
 
 ---
 
+## Recent Additions and Defaults (skill + backend)
+
+- **Perception websocket**: Robot streams frames to `/ws/perception` (binary prefix `0x01` video JPEG, `0x02` audio). Text messages carry `hello`, `turn`, `name_update`. Server soft-fails if face/voice libs are missing; user IDs are created on-demand for stats.
+- **Trivia flow**: Kotlin skill calls `/quiz/question`, `/trivia/turn` (LLM phrasing/feedback), and `/memory/trivia` (stats). Local cache + backend persistence (`trivia_stats` table).
+- **Language handling**: Heuristic EN/NO detector client-side; explicit language pinning via `LanguageManager` with Polly voices (`Kendra-Neural` EN, `Ida-Neural` NO). Backend language hinting in prompts; placeholder server-side lang detect.
+- **URLs/IPs**: Default `BACKEND_URL` in the skill points to laptop IP (override via `BACKEND_URL` env). Robot IP tracked in params. Perception WS URL derives from `BACKEND_URL` (`ws://<host>:8000/ws/perception`).
+- **RAG**: Lightweight BM25 (no vector DB) over `DOCUMENTS_PATH` with preference for `qa_pairs.json` when present; falls back to PDFs/txt. Chunk size 800 / overlap 150.
+- **LLM defaults**: Backend uses Ollama (`llama3.2:latest`) by default; HF/LlamaCpp paths are guarded behind torch/transformers availability. System prompt in `config/settings.py` defines the Kaia persona and strict language policy.
+- **Storage/layout**: Backend defaults to local `.cache` for models/caches/docs/vector store/DB (SQLite `furhat_memory.db`). Tables: `users`, `conversations`, `turns`, `trivia_stats`.
+- **Ingestion helper**: `ingestion/web_ingest.py` fetches PDFs via DuckDuckGo HTML + regex (best-effort), storing into a folder you can set as `DOCUMENTS_PATH`.
+
+
 ## Requirements & Installation
 
 This project uses a hybrid dependency management approach: some dependencies are installed via pip into your environment, and others are managed by Poetry (tracked in the `poetry.lock` file).

furhat_skills/Conversation/build.gradle

@@ -1,6 +1,6 @@
 plugins {
-    id "org.jetbrains.kotlin.jvm" version "1.8.21"
-    id 'com.github.johnrengelman.shadow' version '2.0.4'
+    id "org.jetbrains.kotlin.jvm" version "1.9.22"
+    id "com.github.johnrengelman.shadow" version "8.1.1"
 }
 
 apply plugin: 'java'
@@ -48,6 +48,8 @@ dependencies {
     implementation 'com.furhatrobotics.assets:StandardLibraryCollection:1.2.0'
     // Additional dependencies for HTTP calls, JSON processing, and coroutines
     implementation "com.squareup.okhttp3:okhttp:4.10.0"
+    implementation "com.fasterxml.jackson.core:jackson-annotations:2.15.2"
+    implementation "com.fasterxml.jackson.module:jackson-module-kotlin:2.15.2"
     implementation "org.json:json:20210307"
     implementation "org.jetbrains.kotlinx:kotlinx-coroutines-core:1.6.4"
 }
@@ -74,13 +76,13 @@ shadowJar {
     properties.load(project.file('skill.properties').newDataInputStream())
     def version = properties.getProperty('version')
     def name = properties.getProperty('name')
-    archiveName = "${name}_${version}.skill"
+    archiveFileName.set("${name}_${version}.skill")
+    archiveExtension.set("skill")
 
     manifest {
         exclude '**/Log4j2Plugins.dat'
         exclude '**/node_modules'
     }
     from "skill.properties"
     from "assets"
-    extension 'skill'
 }

furhat_skills/Conversation/gradle/wrapper/gradle-wrapper.properties

@@ -1,5 +1,5 @@
 distributionBase=GRADLE_USER_HOME
 distributionPath=wrapper/dists
+distributionUrl=https\://services.gradle.org/distributions/gradle-8.5-bin.zip
 zipStoreBase=GRADLE_USER_HOME
 zipStorePath=wrapper/dists
-distributionUrl=https\://services.gradle.org/distributions/gradle-6.9.4-bin.zip