Instructions to use mstrasser/jeff-adapter-triage with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use mstrasser/jeff-adapter-triage with PEFT:
Task type is invalid.
- Notebooks
- Google Colab
- Kaggle
Configuration Parsing Warning:In adapter_config.json: "peft.task_type" must be a string
jeff-adapter-triage
Support ticket triage. Routes a support ticket or email to a team and assesses urgency, sentiment and the need for human review.
A LoRA adapter for jeff-base v1.4, Jeff's small open decision model (a fine-tune of Qwen3.5-0.8B). You give it a situation and a question with named options; it returns a calibrated probability for every option from one forward pass, with no generated text to parse.
- Adapter ID:
jeff.support.triage, version 1.4.0; short names:triage - Catalogue page: jeffhub.ai/catalogue/jeff.support.triage
- Code, server and SDKs: github.com/firelex/jeff
- Base it needs:
mstrasser/jeff-baseat revisionv1.4. The engine refuses an adapter trained on another base, so this adapter loads only on that revision.
When to use it
- Each ticket, email or chat message needs a team assignment, urgency rating and human-review decision.
- You want to use your own team list, with a short description of each team.
When not to use it
- You need to write a reply; this adapter returns decisions without generating text.
- The decision depends on facts outside the message, such as order history; include them in the request or handle them in code.
What it decides
What a call sends:
| Field | Required | What it holds |
|---|---|---|
company (context) |
yes | One sentence about the organisation the message is sent to. |
channel (context) |
yes | Where the message came from, for example email, web form, chat, app review, social media reply or phone transcript. |
message (input, fields) |
yes | The ticket or email as received, including subject line, quoted thread and signature. |
jeff.support.triage.route (route, choice)
Instructions: Which team should handle this message? If it covers several issues, choose the team for the most important one.
Labels:
- The caller gives the labels (at most 35), each a key and its text.
- Always offered first:
other: Not for any of these teams
jeff.support.triage.urgency (rate, score)
Instructions: How urgently does this need a response?
Labels:
- 0: No action needed
- 1: Can wait a few days
- 2: Handle within a day
- 3: Handle within hours
- 4: Handle immediately
jeff.support.triage.sentiment (rate, score)
Instructions: How does the customer feel?
Labels:
- 0: Very negative, angry or distressed
- 1: Negative
- 2: Neutral
- 3: Positive
- 4: Very positive
jeff.support.triage.needs_human (requires_escalation, noul)
Instructions: Does this need a person to act on it now, rather than an automatic reply? Answer yes for complaints that could escalate, legal or safety issues, or requests an automatic system cannot resolve.
Labels:
- A yes/no question: the answer is
trueorfalse.
How it was measured
Base: Jeff v1.4; adapter trained 2026-10-25.
| Test | Rows | Accuracy |
|---|---|---|
| the in-house test (generator) | 7,256 | 91.2% |
Accuracy is the share of test rows whose top answer was right. Kinds: public, a public benchmark; real, held-out real documents; generator, an in-house test from the same generator as the training data (it measures less than a public or real test does); mixed, several kinds; unknown, not recorded.
When it is confident (from the calibration set): the threshold is the confidence from which an answer counts as confident, precision the share of confident answers that were right, coverage the share of answers that were confident.
| Function | Threshold | Precision | Coverage | Calibration rows |
|---|---|---|---|---|
route |
0.435 | 95.1% | 100.0% | 367 |
urgency |
0.787 | 95.0% | 59.9% | 367 |
sentiment |
0.687 | 95.0% | 88.0% | 367 |
needs_human |
0.544 | 98.9% | 100.0% | 367 |
How to call it
Install the Python SDK and the engine (see the install guide; until the packages are on PyPI, install them from github.com/firelex/jeff).
The example below is "Wrong jacket size in the parcel". A wrong item to exchange is a matter for the orders team.
On your machine (Jeff.local downloads this adapter and the base from Hugging Face the first time):
import jeff
client = jeff.Jeff.local(["triage"])
decide = client.function("jeff.support.triage.route", context={
"company": "An online shop selling outdoor clothing and camping gear across Europe.",
"channel": "email",
}, labels={
"orders": "Orders, deliveries, returns and exchanges",
"payments": "Payments, refunds and invoices",
"product": "Product questions, sizing advice and stock",
"account": "Customer accounts, passwords and newsletters",
})
decision = decide("Subject: Wrong item in my parcel\nHi, my order 88213 arrived today but the parcel had a large Alpine shell jacket instead of the medium I ordered. Please exchange it and send the medium before Saturday. Thanks, Marta")
print(decision.value, decision.confidence)
Through the Jeff API (an API key from jeffhub.ai):
import os
import jeff
with jeff.Jeff.remote("https://api.jeffhub.ai", token=os.environ["JEFF_API_KEY"], timeout_s=30) as client:
decide = client.function("jeff.support.triage.route", context={
"company": "An online shop selling outdoor clothing and camping gear across Europe.",
"channel": "email",
}, labels={
"orders": "Orders, deliveries, returns and exchanges",
"payments": "Payments, refunds and invoices",
"product": "Product questions, sizing advice and stock",
"account": "Customer accounts, passwords and newsletters",
})
decision = decide("Subject: Wrong item in my parcel\nHi, my order 88213 arrived today but the parcel had a large Alpine shell jacket instead of the medium I ordered. Please exchange it and send the medium before Saturday. Thanks, Marta")
print(decision.value, decision.confidence)
curl -s https://api.jeffhub.ai/v1/decide \
-H "Authorization: Bearer $JEFF_API_KEY" \
-H "Content-Type: application/json" \
-d '{"calls": [{"function": "jeff.support.triage.route", "input": "Subject: Wrong item in my parcel\nHi, my order 88213 arrived today but the parcel had a large Alpine shell jacket instead of the medium I ordered. Please exchange it and send the medium before Saturday. Thanks, Marta", "context": {"company": "An online shop selling outdoor clothing and camping gear across Europe.", "channel": "email"}, "labels": {"orders": "Orders, deliveries, returns and exchanges", "payments": "Payments, refunds and invoices", "product": "Product questions, sizing advice and stock", "account": "Customer accounts, passwords and newsletters"}}]}'
Recorded answer (2026-11-03): orders with confidence 1.0.
To download the weights yourself:
hf download mstrasser/jeff-base --revision v1.4 --local-dir jeff-base
hf download mstrasser/jeff-adapter-triage --revision v1.4 --local-dir adapters/triage
Files
| File | What it is |
|---|---|
adapter_model.safetensors, adapter_config.json |
The LoRA weights and their configuration |
readout.safetensors |
The adapter's readout over the answer codes |
decision_config.json |
Answer codes, temperature, prompt layout and the base it was trained on (the SHA-256 of the base's weights, which the engine checks) |
manifest.yaml |
The adapter's manifest: its functions, labels, fields and calibration, as the SDKs read it |
test.jsonl |
The held-out test set the numbers above come from |
calibration.jsonl |
The calibration set the thresholds come from |
LICENSE |
The Apache License 2.0, for the Qwen3.5-0.8B notice below |
Licence
Adapter licence: Apache-2.0
Qwen3.5-0.8B notice: these weights were modified from Qwen3.5-0.8B by the Jeff project: jeff-base is a fine-tune of Qwen3.5-0.8B, and this adapter was trained on top of it. Qwen3.5-0.8B is Copyright 2026 Alibaba Cloud and licensed under the Apache License, Version 2.0; a copy of that licence is in LICENSE.
Jeff is an independent project, not affiliated with or endorsed by the makers of Qwen.
- Downloads last month
- 123