Donut LoRA Hash Lookup (DEPRECATED)
Resolve a bare SHA256 to a model page
- name
- description
- trigger_words
- model_url
- recommended_weight
Workflows travel as JSON, and a lot of shared JSON references LoRAs by SHA256 hash instead of filename - because hashes are stable across renames, which is exactly why CivitAI keys its whole model registry on them. DonutLoRAHashLookup is the node that turns one of those opaque 64-character strings into a name, description, trigger words, and a model page URL. It's the inverse of the file-based lookup: no file selector, just paste a hash and get answers.
It's DEPRECATED, but this one's a pure fossil with a purpose. The pack kept it registered so saved workflows deserialize, and the unified DonutLoRACivitAILookup node now does the same job as its "Hash" source - its body was copied verbatim into the shared engine, per the source comments. If you're building a new graph, use DonutLoRACivitAILookup with source = Hash; if you have an old graph using this node, it keeps working unchanged.
How it works
Paste the SHA256 into hash (at least 8 characters; it gets trimmed and upper-cased). The engine checks the local CivitAI metadata cache first - this is what makes repeat lookups instant and offline-friendly - then queries the CivitAI API by hash if it isn't cached. api_key is optional and only matters for authenticated requests or heavier rate limits.
The outputs
Five, matching the original node's signature exactly:
name- the model's display namedescription- HTML-stripped and truncated to ~500 charstrigger_wordsmodel_url- the CivitAI page, or a search URL when the hash isn't foundrecommended_weight- a FLOAT, the uploader's suggested strength
No preview_image here - that's a file-mode feature. The engine fills the missing version/preview slots from the shared superset with documented defaults so it can route through the same backend as the newer node.
Install
Same as the whole pack:
cd ComfyUI/custom_nodes
git clone https://github.com/DonutsDelivery/ComfyUI-DonutNodes.git donutnodes
cd donutnodes
python -m pip install -r requirements.txt
or ComfyUI Manager → "DonutNodes," then restart.
The honest caveat
A hash lookup is only as good as CivitAI's index. If the hash isn't in their registry, the node hands you a search URL rather than a dead end - and that search-on-hash trick is genuinely useful, since CivitAI's own search handles bare hashes well. Also note this node has no force_refresh: if you're debugging a stale-cache problem, the newer lookup node or DonutLoRALibrary's "Clear Cache" is where that control lives.
Inputs (2)
| Name | Type | Default | Description |
|---|---|---|---|
| hash | STRING | Paste the SHA256 hash of a model file | |
| api_keyopt | STRING | — |
Outputs (5)
| Name | Type | Description |
|---|---|---|
| name | STRING | — |
| description | STRING | — |
| trigger_words | STRING | — |
| model_url | STRING | — |
| recommended_weight | FLOAT | — |