提交

提交图

124 次代码提交

作者 SHA1 备注 提交日期
Tyler "Nex" Kolody 65a25eedc8 Best configurations across all (non-AVI) formats in a single file compression test. 2025-06-02 17:30:01 -07:00
Tyler "Nex" Kolody 2ad944920f Now that everything's working again with the integrated docker, reworked the ffmpeg to actually use the configs, and the configs to correctly bundle and send the codec parameters in a more usable format. 2025-06-02 16:43:51 -07:00
Tyler "Nex" Kolody 3d296e3de5 Reworked the dockerization to all work from the base encoder. Many things are still hard-coded where they shouldn't be, but everything except AV1 is still massively blown up, and AV1 takes absolute ages (15 minutes for a single file). Next steps will be figuring out why h264/5 are still so bloated, but they all run now. 2025-06-01 22:24:51 -07:00
Tyler "Nex" Kolody 9568abb311 Harmonized everything with the new configs, fixed a few issues or inconsistencies. There are still some significant issues related to getting encoding to work the same way in docker and natively. Right now there's a lot of repetition, poor seperation of concerns, etc. because of the new docker way of doing things. 2025-05-31 20:31:04 -07:00
Tyler "Nex" Kolody 483fe3bce0 Updated the configs to use settings based on SO's recommendations, added a few extra parameters that will need to be propogated to ffmpeg, solidified our scope of supported codecs (mpv4, h265, h264, and soft support for av1) and split the configurations that will most impact compression and performance into seperate dictionaries for each to allow more precise configuration. 2025-05-31 19:41:51 -07:00
Tyler "Nex" Kolody bdfa3a2461 Since h265 encoding is so finicky, created a simple docker container to run the necessary script in. This should allow for anyone to use h265, regardless of their python environment. While it should work on Linux and Mac, it has only been tested using WSL 2025-05-31 18:11:08 -07:00
Tyler "Nex" Kolody 0ea202c4c5 Added the necessary optional requirements for using gemini and claude models, as well as the epub and HTML ingestion 2025-05-30 17:40:09 -07:00
Tyler "Nex" Kolody c2d72c7f2b Removed an errant import. 2025-05-30 17:34:11 -07:00
Tyler "Nex" Kolody 08645ae1ad Made the config defined chunk size and overlap the rightful default for various add_ methods. Added an add_epub method, and add_text now supports html 2025-05-30 17:33:11 -07:00
Tyler "Nex" Kolody 5003f63143 updated the example in interactive.py to use the video file type config 2025-05-30 17:27:30 -07:00
Tyler "Nex" Kolody 43b568c039 Updated the init to work with the LLM client 2025-05-30 17:26:56 -07:00
Tyler "Nex" Kolody 9bc2fde636 Added the video file type config to chat_memory. 2025-05-30 17:26:14 -07:00
Tyler "Nex" Kolody f5e906a9e2 Reworked add_chunk and added helper functions to improve error handling 2025-05-30 17:25:20 -07:00
Tyler "Nex" Kolody db03dcb02f Updated the examples to use the config defined file extension for consistency. Created a file_chat.py script to ingest and speak with use-defined files. 2025-05-30 17:16:36 -07:00
Tyler "Nex" Kolody 91010d6f59 Reworked chat.py and added a new llm_client.py file to allow for the use of multiple different LLM endpoints. The configs have also been expanded to accomodate this. In addition, global configs have been added or moved for chunking and overlap, output file type, and video encoding/compression settings. 2025-05-30 17:10:39 -07:00
Saleban Olow 0a6b67339d Update README.md 2025-05-28 12:31:53 -05:00
saleban olow 887872857e readme clean up 2025-05-28 12:26:02 -05:00
saleban olow 8f9f782380 readme clean up 2025-05-28 12:14:02 -05:00
saleban olow ca750a445b readme clean up 2025-05-28 12:07:26 -05:00
saleban olow 94ea2a593b readme clean up 2025-05-28 11:50:50 -05:00
saleban olow 1fe05a6408 readme clean up v0.1.2 2025-05-28 11:44:03 -05:00
saleban olow e06efcf211 readme clean up 2025-05-28 11:41:51 -05:00
saleban olow a3266336d5 v0.1.2 2025-05-28 11:34:46 -05:00
saleban olow 563c4b53fc clean up 1 2025-05-27 11:02:35 -05:00