Merge pull request #98 from ajcwebdev/refactor

Refactor Whisper and Ollama Docker Configuration

Author: ajcwebdev

.github/Dockerfile

@@ -4,56 +4,95 @@
 # 1) Node base image - Using Debian slim for smaller footprint
 # ---------------------------------------------------
 
+# 1. Use Node 22-slim as base image
 FROM node:22-slim AS base
 
-# Install only required system dependencies
+# 2. Install system dependencies
 RUN apt-get update && apt-get install -y --no-install-recommends \
     ffmpeg git make curl ca-certificates cmake python3 python3-pip \
-    libopenblas-dev g++ build-essential && rm -rf /var/lib/apt/lists/* \
-    && apt-get clean
-
-RUN update-ca-certificates
+    libopenblas-dev g++ build-essential \
+ && rm -rf /var/lib/apt/lists/* \
+ && apt-get clean \
+ && update-ca-certificates
 
+# 3. Set working directory to /usr/src/app
 WORKDIR /usr/src/app
 
-# Install yt-dlp
+# 4. Install yt-dlp
 RUN curl -L https://github.com/yt-dlp/yt-dlp/releases/latest/download/yt-dlp \
-    -o /usr/local/bin/yt-dlp && \
-    chmod a+rx /usr/local/bin/yt-dlp
+    -o /usr/local/bin/yt-dlp \
+ && chmod a+rx /usr/local/bin/yt-dlp
 
-# Install tsx globally
+# 5. Install tsx globally
 RUN npm install -g tsx
 
-# Install whisper.cpp and download models
-RUN git clone --depth=1 https://github.com/ggerganov/whisper.cpp.git && \
-    cd whisper.cpp && \
-    cmake -B build && \
-    cmake --build build -j --config Release && \
-    ./models/download-ggml-model.sh large-v3-turbo && \
-    ./models/download-ggml-model.sh base && \
-    ./models/download-ggml-model.sh tiny && \
-    rm -rf .git
-
-# Copy package files and install deps
+# 6. Clone whisper.cpp at specific tag (v1.7.3)
+RUN git clone --depth=1 https://github.com/ggerganov/whisper.cpp.git
+
+# 7. Configure with CMake (verbose output)
+RUN cmake -B whisper.cpp/build -S whisper.cpp \
+    -DGGML_NATIVE=OFF \
+    -DCMAKE_BUILD_TYPE=Release \
+    -DCMAKE_VERBOSE_MAKEFILE=ON \
+    -DWHISPER_BUILD_EXAMPLES=ON
+
+# 8. Build; if it fails, print CMakeError.log & CMakeOutput.log
+RUN cmake --build whisper.cpp/build --config Release -j \
+ || (cat whisper.cpp/build/CMakeFiles/CMakeError.log && \
+     cat whisper.cpp/build/CMakeFiles/CMakeOutput.log && false)
+
+# 9. Print out the files in the build/bin folder for debugging
+RUN ls -l whisper.cpp/build/bin || true
+
+# 10. Make sure the whisper-cli binary is executable + symlink to /usr/local/bin
+RUN chmod +x /usr/src/app/whisper.cpp/build/bin/whisper-cli \
+ && ln -s /usr/src/app/whisper.cpp/build/bin/whisper-cli /usr/local/bin/whisper-cli
+
+# 11. Confirm that whisper-cli is found
+RUN ls -l /usr/local/bin/whisper-cli
+
+# 12. Download the “base” and “tiny” models
+RUN whisper.cpp/models/download-ggml-model.sh base && \
+    whisper.cpp/models/download-ggml-model.sh tiny
+
+# 12a. Expose environment variables to reference model paths
+ENV WHISPER_BASE_MODEL=/usr/src/app/whisper.cpp/models/ggml-base.bin
+ENV WHISPER_TINY_MODEL=/usr/src/app/whisper.cpp/models/ggml-tiny.bin
+
+# 13. Copy package.json and package-lock.json
 COPY package*.json ./
+
+# 14. Install Node deps
 RUN npm ci --production && npm cache clean --force
 
-# Copy source code
+# 15. Copy source code
 COPY src ./src
+
+# 16. Copy Docker entrypoint script
 COPY .github/docker-entrypoint.sh ./
+
+# 17. Set permissions to make the script executable
 RUN chmod +x /usr/src/app/docker-entrypoint.sh
 
 # ---------------------------------------------------
 # 2) Setup Ollama with models
 # ---------------------------------------------------
 
+# 1. Use the Ollama image as a base
 FROM ollama/ollama:latest AS ollama
+
+# 2. Set working directory to Ollama config directory
 WORKDIR /root/.ollama
 
-# Start Ollama server and pull models
+# 3. Start Ollama server and pull models
 RUN ollama serve & \
     sleep 10 && \
-    ollama pull llama3.2:1b && \
+    ollama pull qwen2.5:0.5b && \
+    ollama ls && \
+    echo "Listing /root/.ollama after qwen2.5:0.5b pull:" && \
+    ls -lh /root/.ollama || true && \
+    echo "Listing /root/.ollama/models after qwen2.5:0.5b pull:" && \
+    ls -lh /root/.ollama/models || true && \
     pkill ollama
 
 # ---------------------------------------------------
@@ -62,20 +101,30 @@ RUN ollama serve & \
 
 FROM base
 
-# Copy Ollama binary and the pre-downloaded models
+# 1. Copy Ollama binary
 COPY --from=ollama /bin/ollama /usr/local/bin/ollama
+
+# 2. Copy pre-downloaded models
 COPY --from=ollama /root/.ollama /root/.ollama
 
+RUN echo "Listing /root/.ollama in final stage:" && \
+    ls -lh /root/.ollama || true && \
+    echo "Listing /root/.ollama/models in final stage:" && \
+    ls -lh /root/.ollama/models || true
+
+# Set environment variables for Whisper
 ENV WHISPER_FORCE_CPU=1
 ENV WHISPER_NO_GPU=1
 
-# Create content directory first
+# 3. Create content directory first
 RUN mkdir -p /usr/src/app/content
 
-# Set proper permissions for the entire app directory including content
-RUN chown -R node:node /usr/src/app && \
-    chmod -R 755 /usr/src/app && \
-    chmod 777 /usr/src/app/content  # Ensure content dir is fully writable
+# 4. Set proper permissions for the entire app directory including content
+RUN chown -R node:node /usr/src/app \
+ && chmod -R 755 /usr/src/app \
+ && chmod 777 /usr/src/app/content \
+ && chown -R node:node /root/.ollama \
+ && chmod -R 755 /root/.ollama
 
 # Switch to non-root user
 USER node

.github/docker-entrypoint.sh

@@ -1,14 +1,37 @@
 #!/bin/sh
 # .github/docker-entrypoint.sh
 
+# Enable error logging
+set -e
+
+log_error() {
+    echo "Error: $1" >&2
+    if [ -f "/tmp/cmake_config.log" ]; then
+        echo "CMake configuration log:" >&2
+        cat /tmp/cmake_config.log >&2
+    fi
+    if [ -f "/tmp/cmake_build.log" ]; then
+        echo "CMake build log:" >&2
+        cat /tmp/cmake_build.log >&2
+    fi
+    exit 1
+}
+
+# Start Ollama server in the background
+echo "Starting Ollama server..."
+ollama serve &
+
+# Wait for Ollama server to start
+sleep 5
+
 # If first argument is "serve", then start the server.
 if [ "$1" = "serve" ]; then
     echo "Starting Autoshow server..."
     # Remove first arg ("serve") so we don't pass that to the server script.
     shift
-    exec tsx --no-warnings --experimental-sqlite src/server/index.ts "$@"
+    tsx --no-warnings --experimental-sqlite src/server/index.ts "$@" || log_error "Server failed to start"
 fi
 
 # Otherwise, run the CLI by default.
 echo "Running Autoshow CLI..."
-exec tsx --no-warnings --experimental-sqlite src/cli/commander.ts "$@"
+tsx --no-warnings --experimental-sqlite src/cli/commander.ts "$@" || log_error "CLI failed to start"

docs/docker.md

@@ -14,7 +14,7 @@ You can run any of the `as` CLI commands by passing arguments to the container v
 npm run docker-cli -- \
   --video "https://www.youtube.com/watch?v=MORMZXEaONk" \
   --whisper base \
-  --ollama "LLAMA_3_2_3B"
+  --ollama
 ```
 
 ### Run the Server with Docker

docs/examples.md

@@ -125,7 +125,7 @@ npm run as -- \
   --whisper base \
   --chatgpt GPT_4_TURBO \
   --prompt summary shortChapters \
-  --noCleanUp
+  --saveAudio
 ```
 
 Here’s what’s happening in this single command:
@@ -137,7 +137,7 @@ Here’s what’s happening in this single command:
 5. **Transcription**: Uses the `--whisper base` model to transcribe each video in a Docker container.
 6. **LLM**: Uses OpenAI ChatGPT’s GPT-4 Turbo model (`--chatgpt GPT_4_TURBO`) to process the transcripts.
 7. **Prompt**: Generates both a summary and short chapter descriptions (`--prompt summary shortChapters`).
-8. **No Clean Up**: Keeps any intermediary or downloaded files around (`--noCleanUp`) so you can inspect them after the run.
+8. **No Clean Up**: Keeps any intermediary or downloaded files around (`--saveAudio`) so you can inspect them after the run.
 
 ### Process Podcast RSS Feed
 
@@ -248,15 +248,15 @@ npm run as -- \
   --speakerLabels \
   --chatgpt GPT_4 \
   --prompt summary longChapters \
-  --noCleanUp
+  --saveAudio
 ```
 
 - **Input**: Process an RSS feed
 - **RSS**: Skip the first 2 items with `--skip 2`
 - **Transcription**: Use AssemblyAI (`--assembly`) with speaker labels (`--speakerLabels`)
 - **LLM**: Use ChatGPT’s GPT-4 model (`--chatgpt GPT_4`)
 - **Prompt**: Request both a summary and long chapters
-- **Utility**: Keep all intermediate files (`--noCleanUp`)
+- **Utility**: Keep all intermediate files (`--saveAudio`)
 
 ## Transcription Options
 
@@ -674,22 +674,35 @@ npm run test-docker
 Benchmark tests, each compare different size models for `whisper.cpp` and a Dockerized version.
 
 ```bash
-npm run bench-tiny
-npm run bench-base
-npm run bench-small
-npm run bench-medium
-npm run bench-large
-npm run bench-turbo
+npx tsx --test test/bench/tiny.test.ts
+npx tsx --test test/bench/base.test.ts
+npx tsx --test test/bench/small.test.ts
+npx tsx --test test/bench/medium.test.ts
+npx tsx --test test/bench/large.test.ts
+npx tsx --test test/bench/turbo.test.ts
+```
+
+Test all available models for a certain LLM service.
+
+```bash
+npx tsx --test test/models/chatgpt.test.ts
+npx tsx --test test/models/claude.test.ts
+npx tsx --test test/models/cohere.test.ts
+npx tsx --test test/models/gemini.test.ts
+npx tsx --test test/models/mistral.test.ts
+npx tsx --test test/models/fireworks.test.ts
+npx tsx --test test/models/together.test.ts
+npx tsx --test test/models/groq.test.ts
 ```
 
 ## Skip Cleanup of Intermediate Files
 
-If you want to keep downloaded or temporary files for debugging or reprocessing purposes, use `--noCleanUp`. This prevents the CLI from deleting intermediary or cached files after finishing its run.
+If you want to keep the downloaded audio file for debugging or reprocessing purposes, use `--saveAudio`. This prevents the CLI from deleting WAV files after finishing its run.
 
 ```bash
 npm run as -- \
   --video "https://www.youtube.com/watch?v=MORMZXEaONk" \
-  --noCleanUp
+  --saveAudio
 ```
 
 ## Chat with Show Notes

package.json

@@ -18,8 +18,8 @@
   "scripts": {
     "tsx:base": "tsx --env-file=.env --no-warnings --experimental-sqlite",
     "setup": "bash ./scripts/setup.sh",
-    "setup-docker": "docker build -t autoshow -f .github/Dockerfile .",
-    "docker-setup": "docker build -t autoshow -f .github/Dockerfile .",
+    "setup-docker": "docker build --progress=plain -t autoshow -f .github/Dockerfile .",
+    "docker-setup": "docker build --progress=plain -t autoshow -f .github/Dockerfile .",
     "setup-all": "npm run setup && npm run docker-setup",
     "as": "npm run tsx:base -- src/cli/commander.ts",
     "video": "npm run as -- --video",
@@ -29,32 +29,20 @@
     "rss": "npm run as -- --rss",
     "info": "npm run as -- --info",
     "serve": "npm run tsx:base -- --watch --experimental-sqlite src/server/index.ts",
-    "test-server-local": "npm run tsx:base -- src/server/tests/fetch-local.ts",
-    "ts": "npm run tsx:base -- src/server/tests/fetch-local.ts",
-    "test-server-all": "npm run tsx:base -- src/server/tests/fetch-all.ts",
+    "clean": "npm run tsx:base scripts/cleanContent.ts",
+    "ta": "tsx --test test/all.test.ts",
+    "test-all": "tsx --test test/all.test.ts",
     "t": "npm run test-local",
-    "bench-tiny": "tsx --test test/bench/tiny.test.ts",
-    "bench-base": "tsx --test test/bench/base.test.ts",
-    "bench-small": "tsx --test test/bench/small.test.ts",
-    "bench-medium": "tsx --test test/bench/medium.test.ts",
-    "bench-large": "tsx --test test/bench/large.test.ts",
-    "bench-turbo": "tsx --test test/bench/turbo.test.ts",
-    "test-models-chatgpt": "tsx --test test/models/chatgpt.test.ts",
-    "test-models-claude": "tsx --test test/models/claude.test.ts",
-    "test-models-cohere": "tsx --test test/models/cohere.test.ts",
-    "test-models-gemini": "tsx --test test/models/gemini.test.ts",
-    "test-models-mistral": "tsx --test test/models/mistral.test.ts",
-    "test-models-fireworks": "tsx --test test/models/fireworks.test.ts",
-    "test-models-together": "tsx --test test/models/together.test.ts",
-    "test-models-groq": "tsx --test test/models/groq.test.ts",
     "test-local": "tsx --test test/local.test.ts",
     "test-docker": "tsx --test test/docker.test.ts",
     "test-services": "tsx --test test/services.test.ts",
-    "test-all": "tsx --test test/all.test.ts",
-    "ta": "tsx --test test/all.test.ts",
-    "clean": "npm run tsx:base scripts/cleanContent.ts",
+    "test-server-all": "npm run tsx:base -- src/server/tests/fetch-all.ts",
+    "ts": "npm run tsx:base -- src/server/tests/fetch-local.ts",
+    "test-server-local": "npm run tsx:base -- src/server/tests/fetch-local.ts",
     "docker-cli": "docker run --rm --env-file .env -v $PWD/content:/usr/src/app/content autoshow",
     "docker-serve": "docker run -d -p 3000:3000 -v $PWD/content:/usr/src/app/content autoshow serve",
+    "docker-debug": "docker run --rm -it --entrypoint sh autoshow -c 'ls -lh /usr/src/app && ls -lh /usr/src/app/whisper.cpp/models && ls -lh /usr/src/app/whisper.cpp/build/bin'",
+    "ollama-debug": "docker run --rm -it --entrypoint sh autoshow -c 'ls -l /usr/local/bin/ollama && ls -lh /root/.ollama'",
     "prune": "docker system prune -af --volumes && docker image prune -af && docker container prune -f && docker volume prune -af",
     "bun": "bun --env-file=.env --no-warnings src/cli/commander.ts",
     "deno": "deno run --allow-sys --allow-read --allow-run --allow-write --allow-env --unstable-sloppy-imports src/cli/commander.ts"

scripts/setup.sh

@@ -57,7 +57,7 @@ else
     check_ollama_server
 
     # Check and pull required models
-    check_and_pull_model "llama3.2:1b" && check_and_pull_model "llama3.2:3b"
+    check_and_pull_model "llama3.2:1b" && check_and_pull_model "qwen2.5:0.5b"
 fi
 
 # Install npm dependencies

src/cli/commander.ts

@@ -14,8 +14,8 @@
 import { argv, exit } from 'node:process'
 import { fileURLToPath } from 'node:url'
 import { Command } from 'commander'
-import { generatePrompt } from '../process-steps/04-select-prompt'
-import { validateAction, validateLLM, validateTranscription, processAction } from '../utils/validate-option'
+import { selectPrompts } from '../process-steps/04-select-prompt'
+import { validateProcessAction, validateLLM, validateTranscription, processAction } from '../utils/validate-option'
 import { l, err, logCompletionSeparator } from '../utils/logging'
 import { envVarsMap } from '../utils/globals'
 import type { ProcessingOptions } from '../types/process'
@@ -66,7 +66,7 @@ program
   .option('--prompt <sections...>', 'Specify prompt sections to include')
   .option('--printPrompt <sections...>', 'Print the prompt sections without processing')
   .option('--customPrompt <filePath>', 'Use a custom prompt from a markdown file')
-  .option('--noCleanUp', 'Do not delete intermediary files after processing')
+  .option('--saveAudio', 'Do not delete intermediary files after processing')
   // Added options to override environment variables from CLI
   /**
    * Additional CLI options to allow passing API keys from the command line,
@@ -120,13 +120,13 @@ program.action(async (options: ProcessingOptions) => {
 
   // If the user just wants to print prompts, do that and exit
   if (options.printPrompt) {
-    const prompt = await generatePrompt(options.printPrompt)
+    const prompt = await selectPrompts({ printPrompt: options.printPrompt })
     console.log(prompt)
     exit(0)
   }
 
   // 1) Validate which action was chosen
-  const action = validateAction(options)
+  const action = validateProcessAction(options, "action")
 
   // 2) Validate LLM
   const llmServices = validateLLM(options)