23 KiB
Backstory
Backstory is an AI Resume agent that provides context into a diverse career narrative. Backstory will take a collection of documents about a person and provide:
- WIP: Through the use of several custom Language Processing Modules (LPM), develop a comprehensive set of test and validation data based on the input documents. While manual review of content should be performed to ensure accuracy, several LLM techniques are employed in the LPM in order to isolate and remove hallucinations and inaccuracies in the test and validation data.
- WIP: Utilizing quantized low-rank adaption (QLoRA) and parameter effecient tine tuning (PEFT,) provide a hyper parameter tuned and customized LLM for use in chat and content creation scenarios with expert knowledge about the individual.
- Post-training, utilize additional RAG content to further enhance the information domain used in conversations and content generation.
- An integrated document publishing work flow that will transform a "Job Description" into a customized "Resume" for the person the LLM has been trained on, incorporating a multi-stage "Fact Check" to reduce hallucination.
While it can run a variety of LLM models, Backstory is currently running Qwen2.5:7b. In addition to the standard model, the chat pipeline also exposes several utility tools for the LLM to use to obtain real-time data.
Internally, Backstory is built using PyTorch 2.6, and Python 3.11 (several pip packages were not yet available for Python 3.12 shipped with Ubuntu Oracular 24.10, which these containers are based on.)
This system was built to run on commodity hardware, for example the Intel Arc B580 GPU with 12G of RAM.
Zero to Hero
Before you spend too much time learning how to customize Backstory, you may want to see it in action with your own information. Fine-tuning the LLM with your data can take a while, so you might want to see what the system can do just by utilizing retrieval-augmented generation.
The ./docs directory has been seeded with an AI generated persona. That directory is only used during development; actual content should be put into the ./docs-prod directory.
Launching with the defaults (which includes the AI generated persona), you can ask things like Who is Eliza Morgan?
If you want to seed your own data:
docker compose down backstory- Remove everything from docs/:
rm -rf docs/* - Put your generic resume in docs/resume/generic[.pdf,.md,.txt,.docx]
- Remove everything from chromadb/:
rm --rf chromadb/* docker compose up backstory -d
WIP
Backstory works by generating a set of facts about you. Those facts can be exposed to the LLM via RAG, or baked into the LLM by fine-tuning. In either scenario, Backstory needs to know your relationship with a given fact.
WIP notes: Right now, it just uses RAG. I'm working on the PEFT+QLoRA code. So take this section as aspirational... (patches welcome)
To facilitate this, Backstory expects the documents it reads to be marked with information that highlights your role in relation to the document. That information is either stored within each document as Front Matter (YAML) or as a YAML sidecar file (a file with the same name as the content, plus the extension .yml)
The two key items expected in the front matter / sidecar are:
---
person:
role:
---
For example, a file resume.md could have the following either as front matter or in the file resume.md.yml:
---
person: James Ketrenos
role: This resume is about James Ketrenos and refers to his work history.
---
A document from a project you worked on, in my case backstory, could have the following front matter:
---
person: James Ketrenos
role: Designed, built, and deployed the application described in this document.
---
During both RAG extraction and during fine-tuning, that context information is provided to the LLM so it can better respond to queries about the user and that user's specific roles.
This project is seeded with a minimal resume and document about backstory. Those are present in the docs/ directory, which is where you will place your content. If you do not replace anything and run the system as-is, Backstory will be able to provide information about me via RAG (there is fine-tuned data provided in this project archive.)
Installation
This project uses docker containers to build. As this was originally written to work on an Intel Arc B580 (Battlemage), it requires a kernel that supports that hardware, such as the one documented at Intel Graphics Preview, which runs in Ubuntu Oracular (24.10)..
NOTE: You need 'docker compose' installed. See Install Docker Engine on Ubuntu
Building
NOTE: You need 'docker compose' installed. See Install Docker Engine on Ubuntu
git clone https://github.com/jketreno/backstory
cd backstory
docker compose build
Containers
This project provides the following containers:
| Container | Purpose |
|---|---|
| backstory | Base container with GPU packages installed and configured. Main server entry point. Exposes an HTTPS entrypoint for use by frontend development |
| backstory-prod | Base container with GPU packages installed and configured. Main server entry point. Exposes an HTTP entrypoint for exposing via nginx or other reverse proxy server. Serves static files generated by frontend. |
| frontend | Frontend development and building static file for backstory-prod. |
| jupyter | backstory + Jupyter notebook for running Jupyter sessions |
| miniircd | Tiny deployment of an IRC server for testing IRC agents |
| ollama | Installation of Intel's pre-built Ollama.cpp |
While developing Backstory, sometimes Hugging Face is used directly with models loaded via PyTorch. At other times, especially during rapid-development, the ollama deployment is used. This combination allows you to easily access GPUs running either locally (via the local ollama or HF code)
To see which models are easily deployable with Ollama, see the Ollama Model List.
Prior to using a new model, you need to download it:
MODEL=qwen2.5:7b
docker compose exec -it ollama ollama pull ${MODEL}
To download many common models for testing against, you can use the fetch-models.sh script which will download:
- qwen2.5:7b
- llama3.2
- mxbai-embed-large
- deepseek-r1:7b
- mistral:7b
To run the script:
docker compose exec -it ollama /fetch-models.sh
The persisted volume mounts (./cache and ./ollama) can grow quite large with models, GPU kernel caching, etc. During the development of this project, the cache directory has grown to consume ~250G of disk space.
Inside the cache you will see directories like:
| Directory | Size | What's in it? |
|---|---|---|
| datasets | 23G | If you download any HF datasets, they will be here |
| hub | 310G | All of the HF models will show up here. docker exec backstory shell "huggingface-cli scan-cache" |
| libsycl_cache | 2.9G | Used by... libsycl. It caches pre-compiled things here. |
| modules | ~1M | Not sure what created this. It has some microsoft code, so maybe from markitdown? |
| neo_compiler_cache | 1.1G | If you are on an Intel GPU, this is where JIT compiled GPU kernels go. If you launch a model and it seems to stall out, watch ls -alt cache/neo_compiler_cache to see if Intel's compute runtime (NEO) is writing here. |
I haven't kept up on pruning out old models I'm not using. Sample output of running the hugging-cli command:
$ docker exec backstory shell "huggingface-cli scan-cache -vvv"
REPO ID REPO TYPE REVISION SIZE ON DISK NB FILES LAST_MODIFIED REFS LOCAL PATH
---------------------------------------------------- --------- ---------------------------------------- ------------ -------- ------------- ---------- ---------------------------------------------------------------------------------------------------------------------------------
Matthijs/cmu-arctic-xvectors dataset 36e87b347a6a70f0420445b02ec40c55556f9ed7 21.3M 1 5 weeks ago /root/.cache/hub/datasets--Matthijs--cmu-arctic-xvectors/snapshots/36e87b347a6a70f0420445b02ec40c55556f9ed7
Matthijs/cmu-arctic-xvectors dataset 5c1297a9eb6c91714ea77c0d4ac5aca9b6a952e5 2.4K 2 5 weeks ago main /root/.cache/hub/datasets--Matthijs--cmu-arctic-xvectors/snapshots/5c1297a9eb6c91714ea77c0d4ac5aca9b6a952e5
McAuley-Lab/Amazon-Reviews-2023 dataset 2b6d039ed471f2ba5fd2acb718bf33b0a7e5598e 25.2G 10 3 weeks ago main /root/.cache/hub/datasets--McAuley-Lab--Amazon-Reviews-2023/snapshots/2b6d039ed471f2ba5fd2acb718bf33b0a7e5598e
yahma/alpaca-cleaned dataset 12567cabf869d7c92e573c7c783905fc160e9639 44.3M 2 2 months ago main /root/.cache/hub/datasets--yahma--alpaca-cleaned/snapshots/12567cabf869d7c92e573c7c783905fc160e9639
IDEA-Research/grounding-dino-tiny model a2bb814dd30d776dcf7e30523b00659f4f141c71 690.3M 8 2 days ago main /root/.cache/hub/models--IDEA-Research--grounding-dino-tiny/snapshots/a2bb814dd30d776dcf7e30523b00659f4f141c71
Intel/neural-chat-7b-v3-3 model 7506dfc5fb325a8a8e0c4f9a6a001671833e5b8e 14.5G 10 3 months ago main /root/.cache/hub/models--Intel--neural-chat-7b-v3-3/snapshots/7506dfc5fb325a8a8e0c4f9a6a001671833e5b8e
Qwen/CodeQwen1.5-7B-Chat model 7b0cc3380fe815e6f08fe2f80c03e05a8b1883d8 14.5G 10 4 weeks ago main /root/.cache/hub/models--Qwen--CodeQwen1.5-7B-Chat/snapshots/7b0cc3380fe815e6f08fe2f80c03e05a8b1883d8
TheBloke/neural-chat-7B-v3-2-AWQ model f3c5e4160e0faecf91ca396558527ba13f1efb72 2.3M 6 2 months ago main /root/.cache/hub/models--TheBloke--neural-chat-7B-v3-2-AWQ/snapshots/f3c5e4160e0faecf91ca396558527ba13f1efb72
TheBloke/neural-chat-7B-v3-2-GGUF model 97de3dbd877a4b022eda57b292d0efba0187ed79 7.5G 3 2 months ago main /root/.cache/hub/models--TheBloke--neural-chat-7B-v3-2-GGUF/snapshots/97de3dbd877a4b022eda57b292d0efba0187ed79
black-forest-labs/FLUX.1-dev model 0ef5fff789c832c5c7f4e127f94c8b54bbcced44 57.9G 29 6 weeks ago main /root/.cache/hub/models--black-forest-labs--FLUX.1-dev/snapshots/0ef5fff789c832c5c7f4e127f94c8b54bbcced44
black-forest-labs/FLUX.1-schnell model 741f7c3ce8b383c54771c7003378a50191e9efe9 33.7G 23 6 weeks ago main /root/.cache/hub/models--black-forest-labs--FLUX.1-schnell/snapshots/741f7c3ce8b383c54771c7003378a50191e9efe9
deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B model ad9f0ae0864d7fbcd1cd905e3c6c5b069cc8b562 3.6G 5 2 months ago main /root/.cache/hub/models--deepseek-ai--DeepSeek-R1-Distill-Qwen-1.5B/snapshots/ad9f0ae0864d7fbcd1cd905e3c6c5b069cc8b562
deepseek-ai/DeepSeek-R1-Distill-Qwen-7B model 916b56a44061fd5cd7d6a8fb632557ed4f724f60 15.2G 7 2 months ago main /root/.cache/hub/models--deepseek-ai--DeepSeek-R1-Distill-Qwen-7B/snapshots/916b56a44061fd5cd7d6a8fb632557ed4f724f60
intel/neural-chat-7b-v3 model 7f6ebc113310e0d2ecc92ae94daeddba5493704d 2.3M 7 2 months ago main /root/.cache/hub/models--intel--neural-chat-7b-v3/snapshots/7f6ebc113310e0d2ecc92ae94daeddba5493704d
intel/neural-chat-7b-v3-3 model 7506dfc5fb325a8a8e0c4f9a6a001671833e5b8e 2.3M 7 2 months ago main /root/.cache/hub/models--intel--neural-chat-7b-v3-3/snapshots/7506dfc5fb325a8a8e0c4f9a6a001671833e5b8e
llmware/intel-neural-chat-7b-v3-2-ov model 7a0a312108b4b9c37c739eb83b592c30c9965eb0 2.3M 5 2 months ago main /root/.cache/hub/models--llmware--intel-neural-chat-7b-v3-2-ov/snapshots/7a0a312108b4b9c37c739eb83b592c30c9965eb0
meta-llama/Llama-3.2-3B model 13afe5124825b4f3751f836b40dafda64c1ed062 9.1M 3 3 weeks ago main /root/.cache/hub/models--meta-llama--Llama-3.2-3B/snapshots/13afe5124825b4f3751f836b40dafda64c1ed062
meta-llama/Llama-3.2-3B-Instruct model 0cb88a4f764b7a12671c53f0838cd831a0843b95 9.1M 3 5 weeks ago main /root/.cache/hub/models--meta-llama--Llama-3.2-3B-Instruct/snapshots/0cb88a4f764b7a12671c53f0838cd831a0843b95
microsoft/Florence-2-base model ceaf371f01ef66192264811b390bccad475a4f02 467.1M 9 2 days ago main /root/.cache/hub/models--microsoft--Florence-2-base/snapshots/ceaf371f01ef66192264811b390bccad475a4f02
microsoft/florence-2-base model ceaf371f01ef66192264811b390bccad475a4f02 2.5M 7 2 days ago main /root/.cache/hub/models--microsoft--florence-2-base/snapshots/ceaf371f01ef66192264811b390bccad475a4f02
microsoft/speecht5_hifigan model 6f01b211b404df2e0a0a20ca79628a757bb35854 50.6M 1 5 weeks ago refs/pr/1 /root/.cache/hub/models--microsoft--speecht5_hifigan/snapshots/6f01b211b404df2e0a0a20ca79628a757bb35854
microsoft/speecht5_hifigan model bb6f429406e86a9992357a972c0698b22043307d 50.7M 2 5 weeks ago main /root/.cache/hub/models--microsoft--speecht5_hifigan/snapshots/bb6f429406e86a9992357a972c0698b22043307d
microsoft/speecht5_tts model 30fcde30f19b87502b8435427b5f5068e401d5f6 585.7M 7 5 weeks ago main /root/.cache/hub/models--microsoft--speecht5_tts/snapshots/30fcde30f19b87502b8435427b5f5068e401d5f6
microsoft/speecht5_tts model a01d4f293234515125d07f68be3c36d739ccac93 585.4M 1 5 weeks ago refs/pr/28 /root/.cache/hub/models--microsoft--speecht5_tts/snapshots/a01d4f293234515125d07f68be3c36d739ccac93
mistralai/Mistral-Small-3.1-24B-Instruct-2503 model 247c7a102f360e2ab181caf6aa7e8144316fd488 96.1G 25 5 weeks ago main /root/.cache/hub/models--mistralai--Mistral-Small-3.1-24B-Instruct-2503/snapshots/247c7a102f360e2ab181caf6aa7e8144316fd488
openlm-research/open_llama_3b_v2 model 4293833c8795656cdacfae811f713ada0e7a2726 6.9G 1 2 months ago refs/pr/16 /root/.cache/hub/models--openlm-research--open_llama_3b_v2/snapshots/4293833c8795656cdacfae811f713ada0e7a2726
openlm-research/open_llama_3b_v2 model bce5d60d3b0c68318862270ec4e794d83308d80a 6.9G 6 2 months ago main /root/.cache/hub/models--openlm-research--open_llama_3b_v2/snapshots/bce5d60d3b0c68318862270ec4e794d83308d80a
openlm-research/open_llama_7b_v2 model e5961def23172a2384543940e773ab676033c963 13.5G 10 3 months ago main /root/.cache/hub/models--openlm-research--open_llama_7b_v2/snapshots/e5961def23172a2384543940e773ab676033c963
runwayml/stable-diffusion-v1-5 model 451f4fe16113bff5a5d2269ed5ad43b0592e9a14 5.5G 15 6 weeks ago main /root/.cache/hub/models--runwayml--stable-diffusion-v1-5/snapshots/451f4fe16113bff5a5d2269ed5ad43b0592e9a14
sentence-transformers/all-MiniLM-L6-v2 model c9745ed1d9f207416be6d2e6f8de32d1f16199bf 91.6M 11 2 months ago main /root/.cache/hub/models--sentence-transformers--all-MiniLM-L6-v2/snapshots/c9745ed1d9f207416be6d2e6f8de32d1f16199bf
stabilityai/stable-diffusion-xl-base-1.0 model 462165984030d82259a11f4367a4eed129e94a7b 7.1G 19 1 day ago main /root/.cache/hub/models--stabilityai--stable-diffusion-xl-base-1.0/snapshots/462165984030d82259a11f4367a4eed129e94a7b
unsloth/Mistral-Small-24B-Base-2501-unsloth-bnb-4bit model 4e277e563e75dc642a9947b0a5e42b16440c9546 15.7G 12 5 weeks ago main /root/.cache/hub/models--unsloth--Mistral-Small-24B-Base-2501-unsloth-bnb-4bit/snapshots/4e277e563e75dc642a9947b0a5e42b16440c9546
Done in 0.0s. Scanned 29 repo(s) for a total of 326.8G.
And inside ollama:
| Directory | Size | What's in it? |
|---|---|---|
| models | 32G | All models downloaded via ollama pull .... Run docker exec ollama ollama list |
Sample output of running ollama list:
$ docker exec ollama ollama list
ggml_sycl_init: found 1 SYCL devices:
NAME ID SIZE MODIFIED
mxbai-embed-large:latest 468836162de7 669 MB 15 hours ago
qwen2.5:3b 357c53fb659c 1.9 GB 10 days ago
mistral:7b f974a74358d6 4.1 GB 2 weeks ago
qwen2.5:7b 845dbda0ea48 4.7 GB 3 weeks ago
llama3.2:latest a80c4f17acd5 2.0 GB 3 weeks ago
dolphin-phi:latest c5761fc77240 1.6 GB 6 weeks ago
llama3.2-vision:latest 085a1fdae525 7.9 GB 6 weeks ago
llava:latest 8dd30f6b0cb1 4.7 GB 6 weeks ago
deepseek-r1:1.5b a42b25d8c10a 1.1 GB 7 weeks ago
deepseek-r1:7b 0a8c26691023 4.7 GB 7 weeks ago
Running
In order to download Hugging Face models, you need to have a Hugging Face token. See https://huggingface.co/settings/tokens for information on obtaining a token.
Edit .env to add the following:
HF_ACCESS_TOKEN=<access token from huggingface>
HF_HOME=/root/.cache
HF_HOME is set for running in the containers to point to a volume mounted directory which will enable model downloads to be persisted.
NOTE: Models downloaded by most examples will be placed in the ./cache directory, which is bind mounted to the container.
Backstory
If you just want to run the pre-built environment, you can run:
docker compose up -d
That will launch all the required containers. Once loaded, the following ports are exposed:
Container: backstory-prod
- 8911 - http for the chat server. If you want https (recommended) then you should use an nginx reverse proxy to provide this endpoint. See src/server.py WEB_PORT and docker-compose
portsunder thebackstoryservice. This port is safe to be exposed to the Internet if you want to expose this from your own service.
Container: backstory
- 8912 - https for the development chat server. Do not expose this port to the Internet. The chat server running on 8912 uses qwen-2.b:3B instead of the larger 7B model. This allows you to run backstory-prod and backstory (for development) on the same GPU without running out of memory.
- 3000 - During interactive development of the frontend, the React is found at this port. By default, static content is served through port 8911. Do not expose this port to the Internet.
Container: jupyter
Do not expose these ports to the Internet
- 8888 - Jupyter Notebook. You can access this port for a Juptyer notebook running on top of the
backstorybase container. - 60673 - This allows you to connect to Gradio apps from outside the container, provided you launch the Gradio on port 60673
.launch(server_name="0.0.0.0", server_port=60673)
Container: ollama
Do not expose these ports to the Internet
- 11434 - ollama server port. This should not be exposed to the Internet. You can use it via curl/wget locally. The
backstoryandjupytercontainers are on the same Docker network, so they do not need this port exposed if you don't want it. See docker-compose.ymlportsunderollama.
Once the above is running, to launch the backstory shell interactively:
docker compose exec --it backstory shell
Jupyter
docker compose up jupyter -d
The default port for inbound connections is 8888 (see docker-compose.yml). $(pwd)/jupyter is bind mounted to /opt/jupyter in the container, which is where notebooks will be saved by default.
To access the jupyter notebook, go to https://localhost:8888/jupyter.
Monitoring
You can run ze-monitor within the launched containers to monitor GPU usage.
docker compose exec backstory ze-monitor --list
Container 5317c503e771 devices:
Device 1: 8086:A780 (Intel(R) UHD Graphics 770)
Device 2: 8086:E20B (Intel(R) Graphics [0xe20b])
To monitor a device:
docker compose exec backstory ze-monitor --device 2
If you have more than one GPU, the device numbering can change when you reboot. You can specify the PCI ID instead of a device number:
docker compose exec backstory ze-monitor --device 8086:e20b
NOTE: The ability to monitor temperature sensors, etc. is restricted while running in a container. I recommend installing ze-monitor on the host system and running it there.
Sample output:
$ ze-monitor --device 8086:e20b --one-shot
Device: 8086:E20B (Intel(R) Arc(TM) B580 Graphics)
Total Memory: 12809404416
Free memory: [# 4% ]
Sensor 0: 38.0C
Sensor 1: 33.0C
Sensor 2: 38.0C
Power usage: 36.0W
-----------------------------------------------------------------------
PID COMMAND-LINE
USED MEMORY SHARED MEMORY ENGINE FLAGS
-----------------------------------------------------------------------
43344 /opt/ollama/ollama-bin serve
MEM: 197795840 SHR: 0 FLAGS:
44098 /opt/ollama/ollama-bin runner --model /root/....threads 8 --no-mmap --parallel 1 --port 42341
MEM: 1006231552 SHR: 0 FLAGS: DMA COMPUTE
85909 /opt/ollama/ollama-bin runner --model /root/....threads 8 --no-mmap --parallel 1 --port 38085
MEM: 3873189888 SHR: 0 FLAGS: DMA COMPUTE
104468 /opt/ollama/ollama-bin runner --model /root/....threads 8 --no-mmap --parallel 1 --port 42505
MEM: 7101763584 SHR: 0 FLAGS: DMA COMPUTE
132740 /opt/backstory/venv/bin/python /opt/backstory...orkers=32 --parent=9 --read-fd=3 --write-fd=