Advanced Allegro.pl Scraper and Monitoring Developer PROJECT_DESCRIPTION: An experienced web scraper will help build a tool for Allegro.pl to fetch and monitor several data points of various products, ensuring high levels of completeness and accuracy. Ke
Budget / Salary$250–750
TypeFreelance project
LocationRemote
Posted2 hours ago
I'm looking for an experienced developer to build a scraper and monitoring tool for Allegro.pl, the largest e-commerce marketplace in Poland.
Allegro's official API does not give access to other sellers' offers, so the data has to be collected from the website. The solution must handle Allegro's anti-bot protection reliably at scale. Please only apply if you have hands-on experience scraping protected websites.
What it needs to do:
fetch offers by category, search phrase, EAN or product ID,
collect full offer data (see data points below),
for each product, record all sellers offering it with their prices,
refresh the data daily and store the history of price and availability changes,
show a change overview (dashboard or report): price up or down, out of stock, offer removed,
export to CSV/XLSX or a database, ideally also via API.
Scale: several tens of thousands of products monitored daily.
Data points:
Mandatory:
offer ID and URL,
product ID / EAN,
title,
price,
availability / stock quantity,
seller name and ID,
delivery cost (cheapest option),
full description (complete text),
all images (full resolution),
product parameters (key-value),
timestamp of the check.
Preferred:
units sold,
Smart badge,
delivery time,
seller rating and number of ratings,
full category path.
Optional:
seller location,
promotions / coupons.
Acceptance criteria:
mandatory fields: at least 95% completeness per daily run,
preferred fields: at least 80% completeness,
optional fields: best effort,
failed records are retried and logged, never skipped silently,
the full daily run finishes within 24 hours,
an error report per run showing fetched, failed and blocked items.
Annotated screenshots of an example offer are available on request.
Budget: please give a price range (min–max) for the build, plus the estimated monthly running costs (server, proxies). I'm open to a paid pilot on a smaller product set first.
In your proposal, please include:
Your experience with protected sites (DataDome, Cloudflare or similar). Examples welcome.
The technology and approach you'd use for Allegro.
Monthly running costs at 5k, 20k and 50k products.
Whether maintenance after changes on Allegro's side is included, and at what rate.
Ogłoszenie 2
Title: Local AI pipeline (ComfyUI) – furniture packshots & interior shots from product photos, batch, colour-accurate
Description:
I need an experienced ComfyUI / Stable Diffusion / Flux specialist to build a local, batch image-generation pipeline for upholstered furniture (sofas, sofa beds, corner sofas) for a UK furniture seller.
Important: no paid AI APIs (Google, OpenAI, Grok etc.). The whole pipeline must run on self-hosted open models on my own hardware, with no per-image cost.
Input: manufacturer photos, usually only 2–3 usable photos per product, often in a different fabric colour than the variant being listed. Missing shots have to be generated and the fabric recoloured to the target colour.
Output sequence per colour variant, per sales platform:
Lifestyle hero: 6 versions per platform, the same sofa in 6 different realistic, well-styled interiors
Packshot front on a white background (1 image)
Packshot side / 3/4 on a white background (1 image)
Sleeping function, sofa bed unfolded, where applicable (1 image)
Storage compartment open, where applicable (1 image)
Detail close-up: fabric texture, stitching, legs or mechanism (1 image)
That makes up to 11 images per colour variant per platform. Infographics are handled separately and are not part of this job.
Strict requirements (not acceptable otherwise):
identical fabric colour and shade across all shots of a variant, including all 6 lifestyle versions,
the product stays true to the original: shape, proportions, cushions, stitching, legs and fabric texture are not altered or "invented",
no visible AI artefacts: no warped lines, melted details or unrealistic shadows,
minimum 2000 px on the longer side, clean white background on packshots.
Volume: about 3,500 SKUs now and about 5,000 later, so the process must be batch-driven with minimal manual work per product.
Deliverables:
a working pipeline installed and configured on my hardware,
batch processing (input folder → output folder with naming per SKU / variant / platform),
a short user guide and handover session.
Budget: please give a price range (min–max). The project starts with a paid test on one product (the full sequence in one colour variant). I have sample source photos ready.
In your proposal, please include:
Examples of your product-consistent generations (packshots, product-in-interior, recolouring).
Which models and workflow you'd use, and how you'd keep colour and product identity consistent.
Minimum and recommended hardware (GPU VRAM, RAM, disk).
Estimated processing time per full sequence (11 images) on the recommended hardware.
Confirmation that the models allow commercial use of the output.
Your price for the paid test.
Allegro's official API does not give access to other sellers' offers, so the data has to be collected from the website. The solution must handle Allegro's anti-bot protection reliably at scale. Please only apply if you have hands-on experience scraping protected websites.
What it needs to do:
fetch offers by category, search phrase, EAN or product ID,
collect full offer data (see data points below),
for each product, record all sellers offering it with their prices,
refresh the data daily and store the history of price and availability changes,
show a change overview (dashboard or report): price up or down, out of stock, offer removed,
export to CSV/XLSX or a database, ideally also via API.
Scale: several tens of thousands of products monitored daily.
Data points:
Mandatory:
offer ID and URL,
product ID / EAN,
title,
price,
availability / stock quantity,
seller name and ID,
delivery cost (cheapest option),
full description (complete text),
all images (full resolution),
product parameters (key-value),
timestamp of the check.
Preferred:
units sold,
Smart badge,
delivery time,
seller rating and number of ratings,
full category path.
Optional:
seller location,
promotions / coupons.
Acceptance criteria:
mandatory fields: at least 95% completeness per daily run,
preferred fields: at least 80% completeness,
optional fields: best effort,
failed records are retried and logged, never skipped silently,
the full daily run finishes within 24 hours,
an error report per run showing fetched, failed and blocked items.
Annotated screenshots of an example offer are available on request.
Budget: please give a price range (min–max) for the build, plus the estimated monthly running costs (server, proxies). I'm open to a paid pilot on a smaller product set first.
In your proposal, please include:
Your experience with protected sites (DataDome, Cloudflare or similar). Examples welcome.
The technology and approach you'd use for Allegro.
Monthly running costs at 5k, 20k and 50k products.
Whether maintenance after changes on Allegro's side is included, and at what rate.
Ogłoszenie 2
Title: Local AI pipeline (ComfyUI) – furniture packshots & interior shots from product photos, batch, colour-accurate
Description:
I need an experienced ComfyUI / Stable Diffusion / Flux specialist to build a local, batch image-generation pipeline for upholstered furniture (sofas, sofa beds, corner sofas) for a UK furniture seller.
Important: no paid AI APIs (Google, OpenAI, Grok etc.). The whole pipeline must run on self-hosted open models on my own hardware, with no per-image cost.
Input: manufacturer photos, usually only 2–3 usable photos per product, often in a different fabric colour than the variant being listed. Missing shots have to be generated and the fabric recoloured to the target colour.
Output sequence per colour variant, per sales platform:
Lifestyle hero: 6 versions per platform, the same sofa in 6 different realistic, well-styled interiors
Packshot front on a white background (1 image)
Packshot side / 3/4 on a white background (1 image)
Sleeping function, sofa bed unfolded, where applicable (1 image)
Storage compartment open, where applicable (1 image)
Detail close-up: fabric texture, stitching, legs or mechanism (1 image)
That makes up to 11 images per colour variant per platform. Infographics are handled separately and are not part of this job.
Strict requirements (not acceptable otherwise):
identical fabric colour and shade across all shots of a variant, including all 6 lifestyle versions,
the product stays true to the original: shape, proportions, cushions, stitching, legs and fabric texture are not altered or "invented",
no visible AI artefacts: no warped lines, melted details or unrealistic shadows,
minimum 2000 px on the longer side, clean white background on packshots.
Volume: about 3,500 SKUs now and about 5,000 later, so the process must be batch-driven with minimal manual work per product.
Deliverables:
a working pipeline installed and configured on my hardware,
batch processing (input folder → output folder with naming per SKU / variant / platform),
a short user guide and handover session.
Budget: please give a price range (min–max). The project starts with a paid test on one product (the full sequence in one colour variant). I have sample source photos ready.
In your proposal, please include:
Examples of your product-consistent generations (packshots, product-in-interior, recolouring).
Which models and workflow you'd use, and how you'd keep colour and product identity consistent.
Minimum and recommended hardware (GPU VRAM, RAM, disk).
Estimated processing time per full sequence (11 images) on the recommended hardware.
Confirmation that the models allow commercial use of the output.
Your price for the paid test.
Apply on Freelancer →
Project sourced from Freelancer.com. Applications happen directly on the original platform — we never collect your data.