diff --git a/.github/workflows/integration-test.yml b/.github/workflows/integration-test.yml index 0d5513f8..79538fed 100644 --- a/.github/workflows/integration-test.yml +++ b/.github/workflows/integration-test.yml @@ -192,6 +192,7 @@ jobs: run: | ./occ app_api:daemon:register --net host manual_install "Manual Install" manual-install http localhost http://localhost:8080 ./occ app_api:app:register context_chat_backend manual_install --json-info "{\"appid\":\"context_chat_backend\",\"name\":\"Context Chat Backend\",\"daemon_config_name\":\"manual_install\",\"version\":\"${{ fromJson(steps.appinfo.outputs.result).version }}\",\"secret\":\"12345\",\"port\":10034,\"scopes\":[],\"system_app\":0}" --force-scopes --wait-finish + ls -la context_chat_backend/persistent_storage/* - name: Scan files, baseline run: | @@ -216,8 +217,8 @@ jobs: - name: Run the prompts run: | - ./occ background-job:worker 'OC\TaskProcessing\SynchronousBackgroundJob' & - ./occ background-job:worker 'OC\TaskProcessing\SynchronousBackgroundJob' & + ./occ background-job:worker 'OC\TaskProcessing\SynchronousBackgroundJob' > worker1_logs 2>&1 & + ./occ background-job:worker 'OC\TaskProcessing\SynchronousBackgroundJob' > worker2_logs 2>&1 & OUT1=$(./occ context_chat:prompt admin "Which factors are taken into account for the Ethical AI Rating?") echo "$OUT1" @@ -259,15 +260,40 @@ jobs: echo "Memory usage during prompt is stable. No memory leak detected." fi - - name: Show logs + - name: Show server logs if: always() run: | cat data/nextcloud.log - echo '--------------------------------------------------' + + - name: Show context_chat specific logs + if: always() + run: | + cat data/context_chat.log + + - name: Show task processing worker logs + if: always() + run: | + tail -v -n +1 worker?_logs || echo "No worker logs" + + - name: Show main app logs + if: always() + run: | cat context_chat_backend/backend_logs || echo "No main backend logs" + + - name: Show main app JSON logs + if: always() + run: | + tail -v -n +1 context_chat_backend/persistent_storage/logs/ccb.log* || echo "No logs in logs directory" + + - name: Show embedding server logs + if: always() + run: | cat context_chat_backend/em_backend_logs || echo "No main backend logs" - echo '--------------------------------------------------' - tail -v -n +1 context_chat_backend/persistent_storage/logs/* || echo "No logs in logs directory" + + - name: Show embedding server JSON logs + if: always() + run: | + tail -v -n +1 context_chat_backend/persistent_storage/logs/em_server.log* || echo "No logs in logs directory" summary: permissions: diff --git a/Dockerfile b/Dockerfile index 0f454bab..cdc77bad 100644 --- a/Dockerfile +++ b/Dockerfile @@ -1,7 +1,7 @@ # SPDX-FileCopyrightText: 2023 Nextcloud GmbH and Nextcloud contributors # SPDX-License-Identifier: AGPL-3.0-or-later # -FROM nvidia/cuda:12.2.2-runtime-ubuntu22.04 +FROM docker.io/nvidia/cuda:12.2.2-runtime-ubuntu22.04 ARG CCB_DB_NAME=ccb ARG CCB_DB_USER=ccbuser diff --git a/Makefile b/Makefile index f10fed40..ccc6a2e5 100644 --- a/Makefile +++ b/Makefile @@ -21,7 +21,7 @@ help: #.PHONY: build-push #build-push: # docker login ghcr.io -# docker buildx build --push --platform linux/arm64/v8,linux/amd64 --tag ghcr.io/nextcloud/context_chat_backend:4.4.1 --tag ghcr.io/nextcloud/context_chat_backend:latest . +# docker buildx build --push --platform linux/arm64/v8,linux/amd64 --tag ghcr.io/nextcloud/context_chat_backend:4.5.0 --tag ghcr.io/nextcloud/context_chat_backend:latest . .PHONY: run run: @@ -34,5 +34,5 @@ run: register: docker exec master-nextcloud-1 sudo -u www-data php occ app_api:app:unregister context_chat_backend --silent || true docker exec master-nextcloud-1 sudo -u www-data php occ app_api:app:register context_chat_backend manual_install --json-info \ - "{\"id\":\"context_chat_backend\",\"name\":\"Context Chat Backend\",\"daemon_config_name\":\"manual_install\",\"version\":\"4.4.1\",\"secret\":\"12345\",\"port\":10034,\"scopes\":[],\"system\":0}" \ + "{\"id\":\"context_chat_backend\",\"name\":\"Context Chat Backend\",\"daemon_config_name\":\"manual_install\",\"version\":\"4.5.0\",\"secret\":\"12345\",\"port\":10034,\"scopes\":[],\"system\":0}" \ --force-scopes --wait-finish diff --git a/README.md b/README.md index fffad95a..50e5298a 100644 --- a/README.md +++ b/README.md @@ -11,11 +11,11 @@ > > The HTTP request timeout is 50 minutes for all requests and can be changed with the `request_timeout` app config for the php app `context_chat` using the occ command (`occ config:app:set context_chat request_timeout --value=3000`, value is in seconds). The same also needs to be done for docker socket proxy. See [Slow responding ExApps](https://github.com/cloud-py-api/docker-socket-proxy?tab=readme-ov-file#slow-responding-exapps) > -> An end-to-end example on how to build and register the backend manually (with CUDA) is at the end of this readme +> An end-to-end example for dev setups on how to build and register the backend manually (with CUDA) is at the end of this readme > > See the [NC Admin docs](https://docs.nextcloud.com/server/latest/admin_manual/ai/app_context_chat.html) for requirements and known limitations. -## Simple Install +## Install Install the given apps for Context Chat to work as desired **in the given order**: - [AppAPI from the Apps page](https://apps.nextcloud.com/apps/app_api) @@ -32,25 +32,26 @@ Install the given apps for Context Chat to work as desired **in the given order* > [!IMPORTANT] > To avoid task processing execution delay, setup at 4 background job workers in the main server (where Nextcloud is installed). The setup process is documented here: https://docs.nextcloud.com/server/latest/admin_manual/ai/overview.html#improve-ai-task-pickup-speed -## Complex Install (without docker) +## Dev Install (without docker) -0. Install the required apps from [Simple Install](#simple-install) other than Context Chat Backend and setup background job workers +0. Install the required apps from [Install](#install) other than Context Chat Backend and setup background job workers 1. `python -m venv .venv` 2. `. .venv/bin/activate` 3. `pip install --upgrade pip setuptools wheel` 4. Install requirements `pip install -r requirements.txt` 5. Copy example.env to .env and fill in the variables -6. Ensure the config file at `persistent_storage/config.yaml` points to the correct config file (cpu vs gpu). If you're unsure, delete it. It will be recreated upon launching the application. The default is to point to the gpu config. -7. Configure `persistent_storage/config.yaml` for the model name, model type and its parameters (which also includes model file's path and model id as per requirements, see example config) +6. Ensure the config file at `persistent_storage/config.yaml` points to the correct config file (cpu vs gpu). If you're unsure, delete it. It will be recreated upon launching the application. +7. Configure `persistent_storage/config.yaml` 8. Setup postgresql externally or use `dockerfile_scripts/pgsql/install.sh` to install it on a Debian-family system. 9. Set the env var `EXTERNAL_DB` or the `connection` key in the `pgvector` config to the postgresql connection string if you're using an external database. 10. Start the database (see `dockerfile_scripts/pgsql/setup.sh` for an example) -11. `./main.py` -12. [Follow the below steps to register the app in the app ecosystem](#register-as-an-ex-app) +11. `./main_em.py` in one terminal/tmux pane +12. `./main.py` in another +13. [Follow the below steps to register the app in the app ecosystem](#register-as-an-ex-app) -## Complex Install (with docker) +## Dev Install (with docker) -0. Install the required apps from [Simple Install](#simple-install) other than Context Chat Backend and setup background job workers +0. Install the required apps from [Install](#install) other than Context Chat Backend and setup background job workers 1. Build the image *(this is a good place to edit the example.env file before building the container)* `docker build -t context_chat_backend . -f Dockerfile` @@ -67,12 +68,12 @@ occ app_api:daemon:register --net host manual_install "Manual Install" manual-in ``` `host` will be `localhost` if nextcloud can access localhost or `host.docker.internal` if nextcloud is inside a docker container and the backend app is on localhost. -If nextcloud is inside a container, `--add-host` option would be required by your nextcloud container. [See example above, pt. 2](#complex-install-with-docker) +If nextcloud is inside a container, `--add-host` option would be required by your nextcloud container. [See example above, pt. 2](#dev-install-with-docker) **2. Register the app using the deploy daemon (be mindful of the port number and the app's version):** ``` occ app_api:app:register context_chat_backend manual_install --json-info \ -"{\"appid\":\"context_chat_backend\",\"name\":\"Context Chat Backend\",\"daemon_config_name\":\"manual_install\",\"version\":\"4.4.1\",\"secret\":\"12345\",\"port\":10034,\"scopes\":[],\"system_app\":0}" \ +"{\"appid\":\"context_chat_backend\",\"name\":\"Context Chat Backend\",\"daemon_config_name\":\"manual_install\",\"version\":\"4.5.0\",\"secret\":\"12345\",\"port\":10034,\"scopes\":[],\"system_app\":0}" \ --force-scopes --wait-finish ``` The command to unregister is given below (force is used to also remove apps whose container has been removed) @@ -167,7 +168,6 @@ Adjust the example.env to your needs so that it fits your environment docker run \ -v ./config.yaml:/app/config.yaml \ -v ./context_chat_backend:/app/context_chat_backend \ - -v /var/run/docker.sock:/var/run/docker.sock \ --env-file example.env \ -p 10034:10034 \ -e CUDA_VISIBLE_DEVICES=0 \ @@ -186,10 +186,6 @@ docker run \ Mounts the context_chat_backend into the docker image - `-v /var/run/docker.sock:/var/run/docker.sock` - - Mounts the Docker socket file from the host into the container. This is done to allow the Docker client running inside the container to communicate with the Docker daemon on the host, essentially controlling Docker and GPU from within the container. - `-v persistent_storage:/app/persistent_storage` Mounts the persistent storage into the docker instance to keep downloaded models stored for the future. @@ -229,11 +225,9 @@ sudo -u www-data php occ app_api:app:register \ --json-info "{\"appid\":\"context_chat_backend\",\ \"name\":\"Context Chat Backend\",\ \"daemon_config_name\":\"manual_install\",\ - \"version\":\"4.4.1\",\ + \"version\":\"4.5.0\",\ \"secret\":\"12345\",\ - \"port\":10034,\ - \"scopes\":[],\ - \"system_app\":0}" \ + \"port\":10034}" \ --force-scopes \ --wait-finish ``` diff --git a/appinfo/info.xml b/appinfo/info.xml index 8f1035b1..2559d54c 100644 --- a/appinfo/info.xml +++ b/appinfo/info.xml @@ -20,7 +20,7 @@ Install the given apps for Context Chat to work as desired **in the given order* Setup background job workers as described here: https://docs.nextcloud.com/server/latest/admin_manual/ai/overview.html#improve-ai-task-pickup-speed ]]> - 4.4.1 + 4.5.0 agpl Anupam Kumar Context Chat @@ -29,13 +29,13 @@ Setup background job workers as described here: https://docs.nextcloud.com/serve https://github.com/nextcloud/context_chat_backend/issues https://github.com/nextcloud/context_chat_backend.git - + ghcr.io nextcloud/context_chat_backend - 4.4.1 + 4.5.0 diff --git a/changelog.md b/changelog.md index 7f753d26..6bb274a7 100644 --- a/changelog.md +++ b/changelog.md @@ -9,6 +9,56 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/) and this project adheres to [Semantic Versioning](http://semver.org/). +## 4.5.0 - 2025-09-23 + +### Changed +- stable release of the 4.5.0-beta series +- bump max NC version to 33 +- store worker logs and present them better (#200) @kyteinsky + +### Fixed +- better parsing of version string for repairs (#209) @kyteinsky +- suppress tokenizer max context length warning (#200) @kyteinsky +- fix nc_py_api to 0.20.2 (#200) @kyteinsky + +### Added +- add app version to file logs (#201) @kyteinsky + + +## 4.5.0-beta.3 - 2025-09-12 + +### Changed +- actually bump max NC version to 33 + + +## 4.5.0-beta.2 - 2025-09-12 + +### Changed +- bump max NC version to 33 + +### Fixed +- let empty but set title headers pass (#211) @kyteinsky + + +## 4.5.0-beta.1 - 2025-09-09 + +### Fixed +- better parsing of version string for repairs (#209) @kyteinsky + + +## 4.5.0-beta.0 - 2025-08-29 + +### Fixed +- suppress tokenizer max context length warning (#200) @kyteinsky +- fix nc_py_api to 0.20.2 (#200) @kyteinsky + +### Changed +- store worker logs and present them better (#200) @kyteinsky + +### Added +- add app version to file logs (#201) @kyteinsky + + ## 4.4.1 - 2025-07-31 ### Fixed diff --git a/context_chat_backend/chain/query_proc.py b/context_chat_backend/chain/query_proc.py index 1e8561df..2d217d29 100644 --- a/context_chat_backend/chain/query_proc.py +++ b/context_chat_backend/chain/query_proc.py @@ -3,12 +3,23 @@ # SPDX-License-Identifier: AGPL-3.0-or-later # import logging +from sys import maxsize as SYS_MAXSIZE from langchain.llms.base import LLM +from transformers import GPT2Tokenizer from ..types import TConfig logger = logging.getLogger('ccb.chain') +TOKENIZER = GPT2Tokenizer.from_pretrained('gpt2') + + +def get_num_tokens(text: str) -> int: + ''' + Returns the number of tokens in the text using the fast GPT2 tokenizer. + ''' + return len(TOKENIZER.encode(text, max_length=SYS_MAXSIZE, truncation=True)) + def get_pruned_query(llm: LLM, config: TConfig, query: str, template: str, text_chunks: list[str]) -> str: ''' @@ -36,8 +47,8 @@ def get_pruned_query(llm: LLM, config: TConfig, query: str, template: str, text_ ) \ or 4096 - query_tokens = llm.get_num_tokens(query) - template_tokens = llm.get_num_tokens(template.format(context='', question='')) + query_tokens = get_num_tokens(query) + template_tokens = get_num_tokens(template.format(context='', question='')) # remaining tokens after the template, query and 'to be' generated tokens remaining_tokens = n_ctx - template_tokens - query_tokens - n_gen @@ -45,7 +56,7 @@ def get_pruned_query(llm: LLM, config: TConfig, query: str, template: str, text_ # If the query is too long to fit in the context, truncate it (keeping the template) if remaining_tokens <= 0: new_remaining_tokens = n_ctx - template_tokens - n_gen - while query and llm.get_num_tokens(query) > new_remaining_tokens: + while query and get_num_tokens(query) > new_remaining_tokens: query = ' '.join(query.split()[:-10]) if not query: @@ -57,7 +68,7 @@ def get_pruned_query(llm: LLM, config: TConfig, query: str, template: str, text_ while text_chunks and remaining_tokens > 0: context = text_chunks.pop(0) - context_tokens = llm.get_num_tokens(context) + context_tokens = get_num_tokens(context) if context_tokens <= remaining_tokens: accepted_chunks.append(context) diff --git a/context_chat_backend/controller.py b/context_chat_backend/controller.py index a1c3ed5f..96965c76 100644 --- a/context_chat_backend/controller.py +++ b/context_chat_backend/controller.py @@ -354,7 +354,7 @@ def _(sources: list[UploadFile]): if not ( value_of(source.headers.get('userIds')) - and value_of(source.headers.get('title')) + and source.headers.get('title', None) is not None and value_of(source.headers.get('type')) and value_of(source.headers.get('modified')) and source.headers['modified'].isdigit() diff --git a/context_chat_backend/logger.py b/context_chat_backend/logger.py index 0ceefe61..79e99aff 100644 --- a/context_chat_backend/logger.py +++ b/context_chat_backend/logger.py @@ -12,8 +12,11 @@ import os from time import gmtime +from dotenv import load_dotenv from ruamel.yaml import YAML +load_dotenv() + __all__ = ['JSONFormatter', 'setup_logging'] LOG_RECORD_BUILTIN_ATTRS = { @@ -62,6 +65,7 @@ def _prepare_log_dict(self, record: logging.LogRecord): "timestamp": dt.datetime.fromtimestamp( record.created, tz=dt.UTC, ).isoformat(), + "version": os.getenv("APP_VERSION", "unknown"), } if record.exc_info is not None: always_fields["exc_info"] = self.formatException(record.exc_info) diff --git a/context_chat_backend/repair/runner.py b/context_chat_backend/repair/runner.py index 7efdb5ed..173bbf41 100755 --- a/context_chat_backend/repair/runner.py +++ b/context_chat_backend/repair/runner.py @@ -16,10 +16,19 @@ def get_previous_version(version_info_path: str) -> tuple[int, bool]: return (0, False) with open(version_info_path) as f: - version_string = f.read() + version_string = f.read().strip() - major, minor, patch = version_string.split('.') - repairs_pending = not (patch.endswith('+') and version_string.rstrip('+') == os.environ['APP_VERSION']) + if not version_string: + return (0, False) + + splits = version_string.split('.') + major = splits[0] + minor = splits[1] if len(splits) > 1 else '0' + + repairs_pending = not ( + version_string.endswith('+') + and version_string.rstrip('+') == os.environ['APP_VERSION'] + ) return (int(major + minor.zfill(3)), repairs_pending) diff --git a/example.env b/example.env index d0185344..f99e2f87 100644 --- a/example.env +++ b/example.env @@ -18,7 +18,7 @@ AA_VERSION=3.0.0 APP_SECRET=12345 APP_ID=context_chat_backend APP_DISPLAY_NAME=Context Chat Backend -APP_VERSION=4.4.1 +APP_VERSION=4.5.0 APP_HOST=0.0.0.0 APP_PORT=10034 APP_PERSISTENT_STORAGE=persistent_storage diff --git a/main_em.py b/main_em.py index a9e07e0b..5b83b788 100755 --- a/main_em.py +++ b/main_em.py @@ -82,6 +82,7 @@ 'model_files', em_conf.llama['model'], ) + logger.debug(f'Trying model path: {em_conf.llama["model"]}') # if the model file is still not found, raise an error if not os.path.isfile(em_conf.llama['model']): diff --git a/pyproject.toml b/pyproject.toml index 2f1f3c1f..4b55093f 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -2,7 +2,7 @@ # SPDX-License-Identifier: AGPL-3.0-or-later [project] name = "context_chat_backend" -version = "4.4.1" +version = "4.5.0" requires-python = ">=3.11" authors = [ { name = "Anupam Kumar", email = "kyteinsky@gmail.com" } diff --git a/requirements.txt b/requirements.txt index d8cfaa2c..d6e4b6d4 100644 --- a/requirements.txt +++ b/requirements.txt @@ -11,7 +11,8 @@ langchain-postgres langchain-unstructured llama_cpp_python msg-parser -nc_py_api +# fix it to the version before niquests +nc_py_api==0.20.2 odfdo odfpy openpyxl