Ocr faster underload but needs to be faster

[akhileshsharma99]

• Initial commit
• added git.sh
• added api examples
• updated requirements.txt
• skeleton: add turbo monorepo structure
• cleanup: package json conflict
• feat: added pr-branch.sh
• cleanup: remove turbo-ex
• cleanup: fixed eslint-rc with @eslin/migrate-config
• feature: added ci action for pnpm
• feature: finish migration of web to solidjs
• admin: testing if review is needed for pr's to be merged
• add hsuky attempt 1
• feat: prettier is working now potentially
• feature: add Rust extraction copy through simple copy
• cleanup: Rust extraction crate compilig
• bugfix: add missing files
• new react base
• fixed lockfile
• Added pages
• feat: added / and /viewer routes
• feat: remove plausible, changed CI names
• Added radix
• cleanup: refactor rust code to remove warnings and use lib
• cleanup: add default run
• Added radix
• Added styled components
• feat: added tailwind
• bugfix: duplication of files from mac
• Updates
• Updates
• Added cyan color scale classes
• Padding passed as prop for button
• Fixed up button a bit
• feat: added in baseline viewer panels
• feature: add utoipa covering health and task routes
• updated tailwind.config.cjs
• feature: beginning to split both things out
• Fixed BetterButton styling and props for active/resting states
• Added onClick prop to Betterbutton
• added more chunk basics
• fix ts error
• Padding passed as prop for button
• Fixed up button a bit
• Updating
• Added header on homepage
• Fixed up stuff for merging
• feature: segmentchunk without the full features
• Progress on homepage
• feature: create_api_key route
• done adding api key
• done adding api key
• version of prefixedapicontroller
• added pdla
• added email check, will return api key
• Completed homepage static setup and phone sizing
• wip: pdf import / dist is out here
• cleanup: make local dev viable
• feature: add diesel migrations
• Reworking segment containers to be accordians
• wip: more pdf-js testing. CI is failing, it is still working
• mkd.py
• chunk and mkd unit test done, need akhil to test
• chunk and mkd unit test done, need akhil to test
• chunking done
• refactored chunking
• Static for viewer is completed with Accordian addition. Starting JSON piping
• local pg
• sending it
• rebased
• server in runnign state
• idk
• done with s3 config
• server running again
• uopdated task_processor
• changed python to fast
• hella conflicts
• storage service removal successful
• JSON and Markdown rendering + json piping completed
• wip: added react-pdf. Going to A/B test possibly
• feat: using react-pdf
• feat: making a pdf now
• renamed dirs
• added docker files
• Added design fixes to several things in viewer and credit buttons on hpmaepage
• bugfix: mdk
• Working in viewer design
• feature: pdf boxes working on small mode
• pdf boxes are more solid
• feature: pdf view with squares works with all screen sizes
• updated rust version
• error building docker
• feature: added libglib2.0-dev
• Updates
• Completed payload creation from uploading components
• Updates
• wip: local docker setup
• pdla yaml created
• chunkmydocs yaml created
• task_processor yaml created
• rrq yaml created
• Added envs to chunkmydocs
• Added envs to task_processor
• pdla yaml created
• chunkmydocs yaml created
• task_processor yaml created
• rrq yaml created
• Added envs to chunkmydocs
• Added envs to task_processor
• added example secret files
• updated example secret files
• removed duplicates
• added cors
• cleanup: fix local dev by correctly spec'ing command for pdla
• Additions from ishaan
• Updates
• Moved things to API key - request to local server hooked up
• fixed issues
• cleanup: chunkmydocs Dockerfile
• feature: add chunkmydocs to docker-compose file
• Updates
• chunkmydocs build successful
• updated docker tag
• fixed dockerfile
• yaml updated
• yaml updated
• updated
• Update lib.rs
• updated docker
• deployed
• Update README.md
• Update README.md
• added ssl imports
• fixed ssl issues
• rembased
• updated task processor
• pyscripts and deployed
• fixed config errors
• fixed config errors
• Create LICENSE
• updated configs
• deploying to prod
• deployment successful
• removed unused file
• Status page static completed
• Complted static status component
• Merging
• colors for pdf boxes
• udpated model to return input file url too
• removed duplicate config
• docker compose keycloak up:
• metering done. hard coded limit to 100 pages
• docker compose keycloak up:
• done
• no expiration works
• metering done. hard coded limit to 100 pages
• more metering
• usage route
• Rebasing with main
• Waiting on PDF being sent with PDF data
• Added loading states
• Semi completed version
• Semi completed version
• Updates
• Starting state lift
• Frontend almost done - need to add Auth and make PDF reader feature complete
• keycloak under construction
• Completed status page with timer
• Completed status with looader
• keycloak up on kube
• Fixed favicon stuff
• added pypi package
• Potential reset - adding shadcn
• Updates
• rm unused code
• Updates
• fixed pdf stuff
• rebased
• idk
• viewer fixed
• new trigger
• Working on it
• Merginf or the day
• keycloak set up completed
• rebase errors
• auth done
• pdf2png
• i hate pnmp
• added .env.example
• Updates
• Small fix on chevron
• Updating and adding auth
• pdf2png dodne
• axios implemented for unified headers
• added pdf2png docker and server
• phi35 service
• added terraform for GPU VM
• working on auth tokens decoding
• udpated triggers and tables
• Added some auth stuff
• fixed up pricing page with auth state
• auth middleware up
• jsonwebtoken was a fail
• token validation succeful
• lol
• added dockers
• phi 35 updateS
• requirements
• requirements.txt
• requirements.txt
• requirements.txt
• requirements.txt
• requirements.txt
• requirements.txt
• requirements.txt
• requirements.txt
• requirements.txt
• requirements.txt
• requirements.txt
• 4 bit quantize
• 4 bit quantize
• 4 bit quantize
• 4 bit quantize - new model
• 4 bit quantize - new model
• added examples
• fix pdf2png test and added examples
• fix pdf2png test and added examples
• phi 3.5 vision not very good :(
• can download json now from header
• added server side rust code to hit pdf2png service
• starting to fuck around with pdf worker
• finally fixed pdf page count reader
• working on it
• Fixed ts checks
• Fixing ts errors
• Added hover for type on segments in pdf - starting on click to go to segment
• Completed pdf overlay color fixes
• completed click segment to scroll to chunk in viewer
• completed click segment to scroll to chunk in viewer
• token validation succeful
• new migrations added
• updated migrations
• updated migrations
• working on it
• v1 of /user
• migrations and user route done
• create task updated
• get task done
• added user id check on get task
• done with migrations
• updated redoc
• Working on it
• Updates
• Updates
• Updates
• Updates
• Updates
• Updates
• updates
• Updates
• Updates
• Updates
• fixed presigned url issue
• Updates
• Updates
• fixed task fetching
• Updates
• Updates
• minor edits
• added user func
• new tables and refactoring
• sdtripe routs
• tryin gto get stripe working
• create customer and intent done
• terraform written
• terraform updated
• terraform for gcp resources completed
• added self-deployment.md
• added self-deployment.md
• updated dockerfiles
• added web docker
• removed grobid url
• removed table ocr url
• fixing deployment
• deploying web
• fixed pyscript models
• updated terraform
• api
• new token
• fixing rrq issues
• task processor
• webhook for stripe
• api
• new token
• task processor
• webhook for stripe
• stripe webhook and local cli testing
• stripe
• stripe
• stripe
• refactor lib
• Add .DS_Store to .gitignore
• new gitignore
• idk
• bearer
• added rrq on vm
• keycloak pgurl updated
• fixing rrq issues
• added rrq on vm
• keycloak pgurl updated
• removed sample pdfs
• models
• fixing git ignore for ds store
• Updates
• Updates
• Updates
• Adding payments from frontend working
• udpating keycloak ingress
• hooking up table_ocr
• pdf splliting fixed
• pdf splliting fixed
• pdf splitting merged
• table ocr rolled
• pdf service
• add pdf service
• pdf service docker updated
• fixed requirements.txt
• deployed succesfully
• rebased
• admin account creation done
• admin account creation done
• removed warnings
• removed redis from terraform
• fixed crontab issues
• deployed and working
• added table ocr to redoc
• idk
• new deploy
• stable
• segments
• updated server
• deployed
• rename
• added get tasks
• /tasks route done
• fixed ts
• added OS pdf conversion tooling and updated third party liscences:
• added Qwen2-VL-7B-Instruct-AWQ support
• adding dev terraform
• samples
• samples
• imports
• 2B
• Refactor chunking logic to handle titles, section headers, and long segments more efficiently. Ignore page headers/footers and ensure logical grouping within target length.
• qwen service works
• wokring on vnet
• working on NAT gateway
• vllm
• vllm
• vllm
• changes
• batched images tset
• batched images test
• batch vllm inference
• finished batch inference
• finished batch inference - added typing
• fixing bugs batch
• fixing bugs batch
• trying out AWQ
• test
• debugging output
• output modles in pyscripts
• finished new output, no output url anymore direct json in response body
• refactored gcp kube
• Flipped
• Updates
• Updates
• Updates
• Updates
• Updates
• Updates
• Updates
• Updates
• Updates
• Updates
• Updates
• Updates
• Updates
• Updates
• Updates
• Updates
• Updates
• Updates
• UIpdates
• Updates
• Updates
• deploying frontend
• deploying frontend with json instead of file url
• added page count and file name:
• added indexes for tasks
• added indexes for tasks
• indexes
• rm un-used code
• rebasing
• updated up.sql
• lfg
• new prod deployment
• Added smart commits and updated .gitignore
• redoc error
• new comments for redoc
• added segment and chunk and segment type to redoc
• deploying redocs bug fixes
• Removed incomplete pages - deployable
• Frontend stable version - cleaned up pages
• silent sign in and sign out
• silent sign in and sign out fixed
• fixed user bug
• Merging
• Status page theme switched
• fixed user bug
• Fixed pdf segments tailwind bug
• pdf bug fixed
• v1 frontend completed
• v1 frontend completed
• v1 of frontend completed
• v1 of frontend completed
• removing console.logs
• fixing assets bug
• pushed changes
• refactoring yaml files
• rm table ocr from redoc
• deploying redoc bug fix
• updating dpeloment
• fixed react img import bug
• "Refactored tasks and added pipeline configuration"
• "Refactored tasks and added pipeline configuration"
• "Refactored task and added pipeline configuration"
• "Refactored task and added pipeline configuration"
• docker update
• new domain deployment successful
• optmized background image
• working on runtime env for vite
• still trying to set env vars
• updated dockerimage
• working on it
• dynamic env works
• updated web deployment
• udpated dockerimage
• deployed to prod
• deploying redoc bug fix
• deploying redoc bug fix
• updating dpeloment
• updating dpeloment
• fixed react img import bug
• fixed react img import bug
• docker update
• docker update
• new domain deployment successful
• new domain deployment successful
• optmized background image
• optmized background image
• working on runtime env for vite
• working on runtime env for vite
• still trying to set env vars
• still trying to set env vars
• updated dockerimage
• updated dockerimage
• working on it
• working on it
• dynamic env works
• dynamic env works
• updated web deployment
• updated web deployment
• udpated dockerimage
• udpated dockerimage
• deployed to prod
• deployed to prod
• refactored
• refactored
• base
• base
• more stripe funcs
• more stripe funcs
• updates to terraform and secrets
• fixing models
• bentoml server running for images
• working on ocr
• working on paddle ocr
• working on paddle ocr
• merging
• bento ml syntax error
• in CUDA hell
• lfg
• trying png files
• image reading error
• suck in deps hell
• suck in deps hell
• optimizing ocr
• adding reading order to ocr
• adding reading order to ocr
• adding reading order to ocr
• paddleocr reading order
• word box
• working on implementing
• task service started
• batching paddle ocr
• table ocr using PPStructure
• table ocr using PPStructure
• visualizing tabe ocr
• visualizing tabe ocr
• table ocr, rows and cols
• table ocr, rows and cols
• deploying new task processor
• deploying new chunkmydocs
• updated structure model
• experimenting with settings
• experimenting with settings
• experimenting with settings
• experimenting with ocr bs structure
• table ocr and ocr working
• adding models
• model seralizatino incorrect
• adding text to the table output
• consolidated ocr results
• consolidated ocr results
• model error fixed
• model error fixed
• model error fixed
• adding latex
• latex bug
• latex bug
• latex bug
• latex bug
• latex bug
• fixing random bug
• latex was mid, trying different model
• latex inference error
• dependencies issues
• removing latex
• removing latex
• image magikc fixes
• pdf permission errors
• pdf permission errors
• pdf permission errors -> updating policy
• updating policy errors
• updating policy errors
• updating policy errors
• isntalled magick 7
• installed magick 7
• working on paddle ocr
• processing segments
• processing segments
• interservice communicatino error
• iamge path error
• segmeents [] after processing
• file path issues
• iamges are base64 strs
• images are base64 strs
• updated crop image function
• debugging
• converted image paths from str to Path
• image cropping issues
• image cropping issues
• image cropping issues
• updating str to bytepath
• fixing path errors
• ocr issues, adding fallback
• ocr issues, adding fallback
• adding multiprocessing
• removing multiprocessing
• removed debug statements
• error management
• using image magick to convert
• fixed paging issue
• increasing segment bbox size
• bug fix
• math mistake
• mergeing
• new pdf2img
• service URL
• service url
• added image magick to startup.sh
• added image magick to startup.sh with openCL
• ppocr has spelling mistakes
• making text detection optiona;
• adding spell check on ocr
• ocr not going brr
• updating pyhton syntax error
• pdla patch
• new pdf to png
• imporoved spell check
• fixed gpu dependencies
• syntax error
• added pdla secret
• udpated project.toml
• udpated project.toml
• udpated project.toml
• udpated project.toml
• adding optional value to pdf to iamg
• updating example
• new pdla patch
• density before file in image magikc
• scale factor added
• scale factor added
• pdla density
• density
• new attempt at pdf page mem error
• new attempt at pdf page mem error
• convert from path
• deploying pdla
• changing pdla default value
• changing pdla default value
• dev gpu terraform
• updated vars for vm
• tf - new
• tf - new
• new compose
• added new docker compose
• added new docker compose
• added new docker compose gpu runtime
• testing without spell check
• added new docker compose gpu runtime
• updated terraform
• new config
• threadlock
• adding multiprocessing for ocr
• increasing max workers
• trying muliprocessing
• pickle error
• pickle error
• no multiporcessing on gpu
• no multiporcessing on gpu
• parrellel gpu processing
• using threadpool executor
• using threadpool executor
• issues with Tensor
• adding threading locks
• pdla maxxed
• multiprocessing cropping issue
• multiprocessing cropping issue
• pydantic error
• merged with main
• pydantic errors
• new models
• output is not in order
• added ocr strategy
• fixed syntax error
• only cropping if needed
• only cropping if needed
• hierarchical content
• iamgemagick gpu support
• iamgemagick gpu support
• syntax error fix
• timing it
• adding gpu acceleration
• using convert
• barely any difference
• adding latex
• adding latex ocr
• fixed sytax error
• img
• formula ocr optional
• added eta
• removing debug logs
• lang set for table ocr
• cleaning up, removed latex ocr (too slow and sucks)
• cleaning up unused deps
• cleaning up unused deps
• ocr issues
• ocr issues
• ocr issues
• ready to be used by the rest of the systems
• connecting teask_processor
• no more mkd
• parallelization increase
• rust implementation done
• stripe
• clean up
• segment_type error
• seralization issue
• seralization issue
• segment_type missing error
• fixed syntax error
• Fixed polling issue for task status viewer
• checking ocr_strat
• checking ocr_strat
• markdown bug
• syntax error
• syntax error
• model hell
• html to mkd
• done with new task processing
• tasks searlization error fixed.
• fixed task count
• new task count
• pdf to image
• extension error
• cropping error
• getting iamges back, not working
• .
• pyscripts
• fixed scaling factor for page
• new compose
• scaling factor issues
• fixed pdla deploy
• added image url stuff :
• added image url stuff
• task creation has image_url
• rm threadlock and run in threadpool in pdla
• image file path fixes
• s3 upload issues
• done with ocr
• get tasks error fixed
• process lock on / route
• Refactored frontend
• Rebased with main succesfully
• Added OCR strategy and Task stats in viewer & status view
• main test
• removed segment density
• Moved dashboard to pages
• Added curl command placeholder for no tasks
• deployed new pdla
• added timing for ocr
• added timing for ocr
• table html to mkd
• table html to mkd
• table html to mkd
• Cleaned up svg errors, added oCR slection on upload form, images optimized
• fixing mkd table issue
• fixing mkd table issue
• adding \n as string literal
• adding \n as string literal
• | is not |?
• doing it on frontend
• doing it on frontend
• base
• frontend markdown resolved
• Rebasing with main
• Added tables markdown css, html viewing options, API key fixes for small screen menu dropdown
• updated google timeout
• list item fixes
• list item ol start from give digit
• clean ul
• migrations
• new migs
• new api
• new api spec
• wiping pg
• updating gpu
• updated gpu
• separated out task workers
• new workers - fast - highquality
• cleaned up pyscripts
• refactoring the models
• syntax error fix
• fixed syntax error
• fixed syntax error
• fixed pyscripts
• new dockers for processors
• new processor names
• cleanup
• Need to add chunking strategy in upload form
• Updating with main
• Added HTML viewing fixes and chunking strategy button on upload form
• No tasks state text prompt added
• task service docker created
• docker updated
• docker updated
• docker updated
• added uv venv
• added uv venv
• new kube
• paddle ocr issue
• created init
• paddle ocr issues, going to build with gpu
• added libgomp1 to dockerfile
• added openCV deps in docker
• updated num of workers
• repushing
• new kube
• unexpected end of data error
• dockerfile updated becuase cudnn error
• new kube
• updated env for docker image
• dockerignore update
• docker image inly copying over essentials
• docker image inly copying over essentials
• fixed COPY
• kube
• deploying new frontend
• task ocr parsed to float
• get content updated
• prints
• idek
• ready for deployment
• deploying new version
• deploying new version
• deploymetn with ocr
• deployment successful
• new cmds
• Fixed bug for API pricing calculator - fast pricing is now correct
• updated api spec
• Fixed billing amount bugs
• better tests
• stripe tested on dev
• invoicing done
• added authorization
• clean up
• rm unused code
• deployed
• token incorrect
• fixed incorrect server url
• rm cron from self deploy
• deployed newest version
• updated pdla
• updated pdla
• added
• added threadpool
• pdla code updated
• clean up services
• pdla fast new deployment
• deploying higher fast throughput
• updated number of connections
• invoicing fixes
• comment purge
• deployed invoicer update
• updated max dims for segments
• updated task
• cloudflare full strict
• Added small fixes for frontend and OCRBBox view on hover + OCRText for OCRBBox on hover
• Small fix - Removed random console.log
• updating queue consumer
• udpaed secrets
• updated deployment file
• Fixed MZ and Safari bugs for upload componenty on landing
• silly
• Removed speed approximations for Segmentation model cards
• file support
• deployed web
• silly
• added pollyfill for promise with resolvers
• libre office integration
• Fixed chunk_length undefined bug
• open api spec
• new venv
• Removed speed approximations for Segmentation model cards
• deployed web
• added pollyfill for promise with resolvers
• Fixed chunk_length undefined bug
• open api spec
• new venv
• sending to task service
• invalid header
• idk headers in pdfs suck
• updates to generate
• qwen
• qwen batch
• qwen batch
• qwen batch
• qwen batch
• qwen fixes
• qwen fixes
• qwen fixes
• qwen batch
• docx
• ppt, docx support added
• lfg
• lfg
• added task
• updated accept list
• merged
• fixed conflicts in qwen
• LLM based ocr for tables added
• updated html logic
• updated cost values
• fixed conflict in process
• fixed circular deps
• fixed circular deps
• updated default value
• removed print statements
• deploying new version with llm table support
• deployed new version with docx, pptx support
• added dropzone type
• added module declaration for react-dropzone
• deployed with LLM table support
• updated num_workers
• updated example kube env
• adding o1 support
• added latex api
• testing o1
• removed o1 stuff
• fixed monthly order for usage and multi file type support
• notices
• Update README.md
• Update LICENSE
• Delete LICENSE
• Create LICENSE
• Create COMMERCIAL_LICENSE
• bug fixes
• prompting
• Fixed free tier rollover on dashboard
• Api dialog width fix
• updated conversion
• Added file support byline
• added ppocr to table with llm
• camelot for tables
• deploying new web and backend
• adding o1 support
• added latex api
• testing o1
• removed o1 stuff
• fixed monthly order for usage and multi file type support
• bug fixes
• prompting
• updated conversion
• added ppocr to table with llm
• camelot for tables
• Small fix in byline for homepage
• adding sliding window to ppocr
• img to table
• img to table
• ocr result fix
• ocr result fix
• fixed pyscripts scaling issues and time logging
• fixed
• Added a lot of phone fixes
• moved file conversion to process, added pdf url to output
• fixed pyscripts and gen signed url
• rm prints
• Removed authentication from phone
• Fixed up pricing page header and styling
• cleanup - unused code
• cleanup - unused code
• default payment method
• new deployment
• cooked
• desc month
• month desc
• reverting
• reverting
• updated example secrets
• pdf_url
• added new logs for debugging
• removed logs
• adding logs to upload task
• newset version deployed
• changed api schema
• implemented new segment models in task
• update cropping
• Implmented new bbox stuff on frontend
• updated rust in docker
• deploying api schema change
• deploying api schema change
• added message for usage limit exceeded
• Using legacy version of pdfjs
• Added posthog analytics
• updated web
• Update README.md
• multi gpu support
• fixed typo for curl command
• fixed typo for curl command
• Update README.md
• Update README.md
• updated readme
• updated readme
• updated readme
• updated readme
• increased chunkr timeout
• increased timeout to 180s
• Updates
• added lazy static to client
• updated terraform
• add timeout for rrq
• server on fire
• server on fire
• lfg
• updating nodes to stop ooming
• seperated ocr from segmentation
• removed ocr lock
• dep hell
• throughput maxxing
• load model for every request with thread locks
• adding telemetry
• ocr more stable