Decision models · API v1

Decision API

Send a document or a JSON record, get back a probability for every allowed answer of every question, in one model pass. No free text to parse. Each use case (inside a pack) fixes the questions and options; you only send the input. With a custom schema you can also ask your own questions.

Base URL · Auth Authorization: Bearer ma_… · Format JSON, UTF-8

Quick start


Response:


Use decision.option for the answer and decision.band to decide what to do with it: act (run automatically), confirm (ask a person to confirm), human (hand over). Bands come from the pack's thresholds, by default act > 0.9, confirm 0.5–0.9, human < 0.5; you can apply your own thresholds to confidence.

Authentication and keys

Subscribe on the plans and the page shows your key once. Send it as a Bearer token. A key has a request rate, a number of requests it may run at once and a monthly token quota, and may be limited to some packs. For agents and bots, make a key of their own from yours (Plans and agent keys). Keep keys on your server; never put them in browser code. A leaked key is revoked and replaced on request.

POST/v1/decisions

Body

fieldtype
packstringPack id, e.g. accounts-payable. See the reference below or GET /v1/packs.
usecasestringUse-case id within the pack, e.g. approval_gate.
inputstring | objectOne document (text) or record (JSON object). Some use cases require JSON fields, listed per use case.
imagesarrayOptional: 1-8 images as data URLs, for use cases that read images (marked image in the reference). See Images and video.
videoobjectOptional: {"frames": [data URL, …]}, 2-8 frames of a clip in time order, for use cases that read video. See Images and video.
inputsarrayInstead of input: up to 32 inputs decided in one call. Results come back in the same order.
questionsobjectOptional custom schema instead of pack/usecase, see Custom questions.
modelstringOptional; leave it out to use the default. GET /v1/models lists what is available.

Response

field
idRequest id (dec_…), quote it in support requests.
modelModel tier that decided.
results[]One per input. decision: the use case's main question with option, confidence, band. answers: every question with option, confidence, band and probabilities for all options; score questions add expected (probability-weighted level) and level. Ranking use cases return ranking[] of {id, score} instead.
usageinput_tokens, output_tokens (always 0, see Billing), decisions (inputs × questions).
latency_msTime spent on the server.

Images and video

The model reads images as well as text. Send them in images (1-8 images, each a data URL: data:image/jpeg;base64,…; JPEG, PNG, WebP or GIF). A text input can go with them as context, or be left out. Use cases that read images are marked image or video in the reference; the vision pack has seven (document photos, photo quality, screenshots, charts, motion and status lights). With custom questions images are always allowed.

Video goes in as frames: your side extracts 2-8 frames from the clip in time order and sends {"video": {"frames": [...]}}. The model reads them as one clip (what changed, which way something moved, how a state ended). Frames are cheap to take: evenly spaced, with the clip's start and end covered. Some use cases need a clip, and refuse a single image.

limit
images or frames per request8, for one input (media cannot be combined with inputs; send one request per photo or clip)
size3 MB per image, 9 MB per request (the public demo: 5 MB). Downscale before sending: the longest side at 1024 px or less is plenty and cheaper; a frame of a clip at 640-896 px.
costAn image costs input tokens like text: roughly 250 tokens at 640 × 400, 600 at 896 × 504, up to 2,000 for a large photo. 8 frames of 896 px are about 5,000. Counted in usage.input_tokens, with usage.images giving the number of images.

Request


Frames from a video, in the browser (JavaScript)


Frames from a video, on a server (ffmpeg and Python)



Custom questions

Send questions instead of pack/usecase. Each question has a type: choice (named options), score (ordered levels, low to high) or noul (yes/no), an instructions sentence, and criteria: for choice an object {option: "when to pick it"}, for score a list of level descriptions. Write criteria as decision rules; the model reads them.



Other endpoints

GET/v1/models – available tiers and the default.
GET/v1/packs – every pack and use case with its input description, questions and options (no key needed).
GET/v1/packs/{pack} – one pack.
GET/v1/usage – this month's requests, decisions and tokens for your key, plus the plan, its quota and what is left, and the rate limit. GET/v1/keys, POST/v1/keys, DELETE/v1/keys/{prefix} – agent keys.

Errors

Errors return {"error": {"code", "message"}}.

statuscodemeaning
400invalid_json, missing_input, invalid_inputs, invalid_schemaFix the request.
401missing_key, invalid_keyNo key, unknown or revoked key.
403pack_not_allowedThe key is not enabled for this pack.
400unknown_modelThe model is not available here; see /v1/models.
404unknown_pack, unknown_usecaseCheck ids in /v1/packs.
413input_too_largeAn input is over 20,000 characters; split it.
422input_mismatchThe use case needs JSON fields, an image or a clip that the request does not have, or does not read images (text only).
400invalid_mediaNot a data URL, not a real image file, more than 8 images, images together with video, or media together with inputs.
413media_too_large, body_too_largeDownscale the images (longest side 1024 px), or send fewer frames.
501media_unsupportedThis deployment does not read images.
402payment_requiredThe payment for the plan failed, so the key is paused; it resumes by itself when the payment goes through. Update the card in the billing portal (link in Stripe's email) or ask a person to pay: POST /api/checkout gives a link.
429rate_limited, too_many_concurrent, quota_exceeded, agent_budget_exceededWait for Retry-After seconds. quota_exceeded: the plan's tokens are used up until the 1st, or move to a bigger plan. agent_budget_exceeded: this agent key spent its own monthly budget.
403plan_key_requiredOnly the key of a plan can manage agent keys.
409agent_key_limitThe plan's agent keys are all in use; revoke one.
503 / 504busy, timeoutRetry with backoff.

Plans and agent keys

planpriceinput tokens a monthrequests a minuteat onceagent keys
Developer€19 / month100 M (about 250,000 decisions)12025
Team€99 / month1,000 M (about 2.5 million decisions)600425

Prices in euro without VAT. Pay with Stripe on the plans; the page shows your key once. The service runs one model, fast; GET /v1/models shows it. At the monthly limit the API answers 429 quota_exceeded until the 1st and nothing is charged on top. Change or cancel the plan, update the card and download invoices in Stripe's billing portal: POST /api/portal with your plan key returns the link. Business and private deployments are agreed individually: contact.

One plan, many agents. The plan key makes a key for each agent or bot. Agent keys spend the plan's tokens and run under the plan's rate and concurrency; each may also have a monthly monthly_tokens budget of its own, so a runaway agent cannot burn the whole plan, and each can be revoked alone. An agent key cannot make keys.

curl -X POST https://decision.mediaatlas.si/v1/keys -H "Authorization: Bearer ma_..." -H "Content-Type: application/json" \
  -d '{"name": "support-triage-agent", "monthly_tokens": 5000000, "packs": ["core"]}'
# 201 {"key": "ma_...", "prefix": "ma_k3Xf...", "name": "support-triage-agent", "budget_tokens": 5000000, "rpm": 120, "note": "..."}   (the key is shown once)
curl https://decision.mediaatlas.si/v1/keys -H "Authorization: Bearer ma_..."                     # your agent keys, with tokens used
curl -X DELETE https://decision.mediaatlas.si/v1/keys/ma_k3Xf... -H "Authorization: Bearer ma_..."  # revoke one (or POST /v1/keys/<prefix>/revoke)

GET /v1/usage shows the plan, the plan's tokens used and left this month, and for an agent key its own budget. An agent that needs a plan can ask for a payment link: POST /api/checkout with {"plan": "developer"} or "team" returns {"url": "https://checkout.stripe.com/..."} to give to a person. The agent skill wraps all of this.

Billing and limits

Usage is metered per key in input tokens: the input plus the use case's questions and options as the model reads them. Decision models do not generate text, so output_tokens is always 0. Images and video frames count as input tokens too (see Images and video). Longer inputs, more questions and more images cost more tokens; a batch (inputs) costs the same as separate calls but is faster. usage.input_tokens in every response and GET /v1/usage show what was counted. The plan's quota is shared by the plan key and its agent keys (see Plans and agent keys). Request content is not stored, only counts and timings.

Client examples

Python


JavaScript (server side)


C# (.NET)



Pack reference

Generated from the pack definitions served by GET /v1/packs, so it always matches the running server.

Core decisions core

The nine building blocks: routing, guardrails, tool gating, triage, reranking, evals, labeling, real-time control and confidence gates.

Model routing core / routing

Route each prompt to the cheapest model that can answer it well.

Any text: the prompt a user would send to an LLM.

Guardrails core / guardrails

Inputs to a support assistant for an online electronics shop. Block injection, abuse and off-policy requests before they reach the LLM.

Any text: a message about to be sent to a shop-support assistant.

Tool-call gating core / tool_gating

A finance/support agent proposes tool calls. Allow safe ones, ask a human for risky ones, deny the dangerous ones.

JSON {"user_request", "tool_call", "context"?} - the tool call an agent wants to run. Plain text also works (read as a tool call).

Inbox triage core / inbox

Sort a company inbox into reply now, later or archive.

Any text, best as an e-mail with From: / Subject: lines.

Reranking core / reranking

Score ten help-centre passages against three customer queries and reorder them by relevance.

A search query as text (or JSON {"query"}). It is scored against the 10 help-centre passages in packs/<pack>/5-reranking.json.

LLM evals core / evals

Judge assistant answers against the source passage on a 1-5 scale.

JSON {"context", "question", "answer"} - the answer is graded 1-5 against the context. Plain text is skipped.

Bulk labeling core / bulk_labeling

Label product reviews (several languages) with sentiment, topic and language in one pass per row.

Any text, best a product review in EN, SL, DE or HR.

Real-time control core / realtime

Vaccine fridge controller: one decision per sensor tick, latency measured against a 300 ms budget.

JSON sensor tick {"temp_c", "trend_c_per_min", "door", "door_open_min"}, or press "Start simulated stream". Plain text is skipped.

Confidence gate core / confidence_gate

Support tickets, some clear and some ambiguous. Act when confident, confirm in the middle, hand to a human when unsure.

Any text: a customer support ticket. The intent's confidence picks act / confirm / human.

Accounts payable accounts-payable

For finance teams and bookkeeping bureaus: sort incoming documents, catch duplicate invoices, decide approve / hold / escalate and code costs with VAT treatment.

Document type accounts-payable / document_type

Sort the AP inbox: invoice, credit note, payment reminder, quote, receipt or not a payable document, in EN, SL, DE and HR.

OCR text of the first page of an incoming document (or the e-mail body), any of EN, SL, DE, HR.

Duplicate invoices accounts-payable / duplicate_check

Compare two invoice headers and decide whether they are the same bill received twice, with look-alike recurring and instalment invoices as traps.

JSON {"a": header, "b": header} where each header has supplier, vat_id, invoice_no, date, amount, iban and a short description. Plain text is skipped.

Approve, hold or escalate accounts-payable / approval_gate

Three-way match plus fraud checks: approve clean invoices, hold mismatches for the buyer, escalate bank-detail changes and other control risks.

JSON invoice check {"supplier", "supplier_since", "invoice_no", "amount", "po", "goods_receipt", "tolerance", "bank_account", "approver_limit", "notes"}. Plain text is skipped.

Cost coding accounts-payable / cost_coding

Code each invoice line in one pass: cost category, VAT treatment and cost centre.

One invoice line or short invoice summary: supplier (with country), what was bought, VAT note. The buyer is a VAT-registered company in Slovenia.

Agent safety agent-safety

For teams that run AI coding and ops agents: decide whether a tool call, shell command or code change may proceed.

Tool-call firewall agent-safety / tool_call_firewall

Allow, ask or deny a tool call an AI agent wants to make, based on scope, reversibility and secrets exposure.

JSON object {tool, args, context}: the tool call the agent proposes and where it runs (workspace path, environment, what the user asked).

Shell command risk agent-safety / shell_command_risk

Rate a single shell command an agent wants to run as safe, needs review, or destructive.

One shell command as plain text, as the agent would run it from a project directory.

Secret in diff agent-safety / secret_in_diff

Scan a code diff an agent wants to commit for real secrets, tell them from test values and hashes, and pick block, warn or pass.

A unified diff snippet or a few added lines of code or config that an agent is about to commit.

Human handoff agent-safety / human_handoff

Decide from an agent's own status text whether it should keep going, ask the user a question, or stop and hand over.

A short status text from a coding or ops agent: what it just did, what it is about to do, or where it is stuck.

Clinical intake clinical-intake

Synthetic patient messages (EN, SL) triaged for urgency, routing, red flags and clinician review. Decision support only, not a medical device: a clinician owns every clinical decision.

Urgency triage clinical-intake / urgency

Grade a patient message from routine to emergency now (synthetic, decision support only).

Any text: a message a patient or relative sends to a practice (portal, e-mail, SMS), in English or Slovenian.

Role routing clinical-intake / role_routing

Send each non-emergency patient message to the role that should handle it: nurse, GP, specialist team, pharmacist or front desk.

Any text: a non-emergency patient message in English or Slovenian. Emergencies are caught by the urgency use case first.

Red-flag labels clinical-intake / red_flags

Label each message with its red-flag category, who it is about and its language in one pass.

Any text: a patient or relative message in English or Slovenian.

Auto-reply gate clinical-intake / auto_reply_gate

Send a template reply only to purely administrative messages; anything clinical, mixed or unclear goes to a clinician.

Any text: a patient message. The confidence of the handling decision picks act (auto-reply) / confirm (staff checks the draft) / human (clinician).

Cold-chain monitoring cold-chain

For pharmacies, hospitals, labs and food logistics: triage temperature excursions, spot faulty sensors, decide what happens to the stock and who gets woken up.

Excursion triage cold-chain / excursion_triage

Medicine fridge stream: classify every sensor tick as normal, pre-alarm, door alarm, excursion or sensor fault within a 300 ms budget.

JSON sensor tick {"t", "temp_c", "trend_c_per_min", "door", "door_open_min"}, or press "Start fridge stream". Plain text is skipped.

Sensor fault diagnosis cold-chain / sensor_fault

Look at a short reading series and tell a real excursion from a stuck, drifting, glitching or disconnected probe.

JSON {"unit", "interval_min", "readings": [temps in C, oldest first], optional "reference" (second probe or handheld check), "events"}. Plain text is skipped.

Excursion impact cold-chain / excursion_impact

After an excursion: release the stock, quarantine it for QA, or discard it, from peak, duration, stability budget and mean kinetic temperature.

JSON excursion record {"product", "label_storage", "peak_c" or "min_c", "duration_h", "stability_allowance", "cumulative_before_h", optional "mkt_30d_c", "freeze_sensitive", "data"}. Plain text is skipped.

Alarm escalation cold-chain / alarm_escalation

Pick the notification level for an open alarm: log it, notify the site, page the on-call person, or escalate to the responsible person.

JSON alarm {"alarm", "unit", "value", "open_min", "local_time", "staff_on_site", "acknowledged", "stock"}. Plain text is skipped.

Outreach & content content-growth

For sales, marketing and community teams: classify outreach replies, grade posts, flag low-quality AI text and score lead fit.

Outreach reply content-growth / outreach_reply

Classify the intent of an e-mail reply to a cold outreach message so the right follow-up (or none) is triggered.

Any text: the body of an e-mail reply received after a cold outreach message.

Post grade content-growth / post_grade

Grade a short social post or blog intro from 1 to 5 on specificity, a concrete claim and a clear point.

Any text: a short social post, newsletter teaser or the first paragraph of a blog article.

AI slop check content-growth / ai_slop

Flag generic, low-effort filler text (often machine-written) versus specific, opinionated human writing.

Any text: a short paragraph, post, comment or product blurb to check for generic filler.

Lead fit content-growth / lead_fit

Score an inbound or prospected lead against the ideal customer profile: hot, warm or cold.

JSON {"company", "role", "size", "industry", "signal"} - the lead. The confidence picks act / confirm / human.

Content moderation content-moderation

Trust and safety teams decide what stays up: action on posts, policy labels, appeal gate and marketplace listing checks.

Post action content-moderation / post_action

Decide whether a user post stays up, goes to human review, or is removed, with the main policy reason.

Any text: a user-generated post or comment to moderate.

Policy labels content-moderation / policy_labels

Tag a post with the policy category it falls under and whether it targets a specific person.

Any text: a user-generated post or comment.

Appeal gate content-moderation / appeal_gate

Judge an appeal against a removal: auto-restore clear mistakes, keep clear violations removed, send the rest to a human reviewer.

JSON {"content", "original_reason", "appeal_text"} - the removed post, why it was removed, and the user's appeal. The confidence picks act / confirm / human.

Marketplace listing check content-moderation / marketplace_listing

Screen a marketplace listing: prohibited item, likely counterfeit, or OK to publish.

Any text: a marketplace listing title and description.

Contract review contract-review

Customer-side review of supplier contract clauses: label each clause, score its risk against a playbook, find the clause that answers a question and decide accept, negotiate or reject.

Clause labeling contract-review / clause_labels

Label each clause with its type, the party it favours and whether it carries a deadline or an amount.

Any text: one contract clause between a Customer and a Supplier (EN or SL).

Clause risk contract-review / clause_risk

Score a supplier's clause 1-5 against the Customer's negotiation playbook.

Any text: one clause proposed by a Supplier. It is scored against the playbook stated in the question instructions.

Clause retrieval contract-review / clause_retrieval

Find the clauses of a services agreement that answer a reviewer's question and rank them first.

A question about the contract as text (or JSON {"query"}). It is scored against the 11 clauses of a synthetic master services agreement.

Deviation from standard contract-review / deviation

Compare the counterparty's clause with our standard clause and decide: accept, negotiate or reject.

JSON {"standard", "clause"}: our standard position and the clause the counterparty proposes. Plain text is skipped.

Osnovne odločitve (SL) core-sl

Slovenska zrcalna različica paketa core: istih devet gradnikov z enakimi primeri in oznakami, za neposredno primerjavo istega modela v angleščini in slovenščini.

Usmerjanje modelov core-sl / routing

Vsak poziv usmeri k najcenejšemu modelu, ki nanj še dobro odgovori.

Poljubno besedilo: poziv, ki bi ga uporabnik poslal jezikovnemu modelu.

Varovalke core-sl / guardrails

Sporočila za asistenta podpore spletne trgovine z elektroniko. Prestrezi vrivanje navodil, žalitve in zahteve zunaj pravil, preden pridejo do jezikovnega modela.

Poljubno besedilo: sporočilo, ki bo poslano asistentu podpore trgovine.

Nadzor klicev orodij core-sl / tool_gating

Agent za finance in podporo predlaga klice orodij. Varne dovoli, pri tveganih vprašaj človeka, nevarne zavrni.

JSON {"user_request", "tool_call", "context"?} – klic orodja, ki ga želi izvesti agent. Deluje tudi navadno besedilo (prebere se kot klic orodja).

Razvrščanje pošte core-sl / inbox

Razvrsti podjetniški poštni predal: odgovori takoj, pozneje ali arhiviraj.

Poljubno besedilo, najbolje e-sporočilo z vrsticama From: / Zadeva:.

Prerazvrščanje core-sl / reranking

Oceni deset odlomkov iz centra za pomoč glede na tri vprašanja strank in jih razvrsti po ustreznosti.

Iskalna poizvedba kot besedilo (ali JSON {"query"}). Oceni se glede na 10 odlomkov iz centra za pomoč v packs/<pack>/5-reranking.json.

Vrednotenje jezikovnih modelov core-sl / evals

Oceni odgovore asistenta glede na izvorni odlomek na lestvici od 1 do 5.

JSON {"context", "question", "answer"} – odgovor se oceni od 1 do 5 glede na kontekst. Navadno besedilo se preskoči.

Množično označevanje core-sl / bulk_labeling

Ocene izdelkov (v več jezikih) v enem prehodu na vrstico označi s sentimentom, temo in jezikom.

Poljubno besedilo, najbolje ocena izdelka v slovenščini, angleščini, nemščini ali hrvaščini.

Krmiljenje v realnem času core-sl / realtime

Krmilnik hladilnika za cepiva: ena odločitev na odčitek senzorja, zakasnitev merjena glede na proračun 300 ms.

JSON odčitek senzorja {"temp_c", "trend_c_per_min", "door", "door_open_min"} ali pritisnite "Start simulated stream". Navadno besedilo se preskoči. Vrednost "door" je "open" ali "closed".

Prag zaupanja core-sl / confidence_gate

Zahtevki za podporo, nekateri jasni, drugi dvoumni. Ukrepaj, ko si prepričan, v vmesnem območju zahtevaj potrditev, ob negotovosti predaj človeku.

Poljubno besedilo: zahtevek stranke za podporo. Zaupanje v namen določi ukrepaj / potrdi / človek.

Data flywheel data-flywheel

Curate and judge LLM training data: triage rows, pick DPO preferences, grade answers against sources and label rows for the mix.

Row triage data-flywheel / row_triage

Keep, fix or drop an instruction/response pair before it goes into an SFT mix.

JSON {"instruction", "response"} - one supervised fine-tuning row. Plain text is skipped.

DPO preference pairs data-flywheel / preference_pairs

Pick the better of two responses for preference training. Confident pairs go straight into the DPO set, unsure ones to an annotator.

JSON {"prompt", "a", "b"} - one prompt and two candidate responses. The confidence of the preference picks act / confirm / human.

Grounded answer judge data-flywheel / answer_judge

Grade a model answer 1-5 against its source passage and flag hallucinations before it enters the eval scoreboard.

JSON {"context", "question", "answer"} - the answer is graded 1-5 using only the context as ground truth. Plain text is skipped.

Near-duplicate check data-flywheel / near_duplicate

Decide whether two training prompts are duplicates, paraphrases or genuinely different, so the mix is deduplicated without collapsing real variety.

JSON {"a", "b"} - two prompts (or instruction texts) from the training pool. Plain text is skipped.

Dev & ops triage dev-ops

For SRE and platform teams: triage log lines, vet dependencies, classify CI failures and route issues.

Log triage dev-ops / log_triage

Decide whether a log line needs a human now, belongs in the daily digest or can be resolved as noise.

One log line or a short excerpt from a service, host or orchestrator, with timestamps and service names.

Package audit dev-ops / package_audit

Vet a dependency install event: fine, a lookalike of a popular package, or suspicious on its own.

A package install event as a JSON object with name, version, registry, weekly_downloads, age_days, maintainer_count and has_install_script.

CI failure cause dev-ops / ci_failure

Classify a failed CI job as a real failure of the change, a flaky test or an infrastructure problem.

A CI log excerpt as text: job name, the failing step output, and sometimes what the commit changed.

Issue routing dev-ops / issue_routing

Label an incoming issue with its type, the area of the stack it belongs to and a priority.

An issue title and body as free text, in English, Slovenian or German.

Education and tutoring education

Tutors and teachers of school-age learners grade answers, spot misconceptions, decide when to give a hint and read the mood of a message.

Rubric grading education / rubric_grade

Grade a learner's answer against a rubric on a 1-5 scale, for programming, robotics and maths tasks.

JSON {"question", "rubric", "answer"} - the learner's answer graded 1-5 against the rubric. Plain text is skipped.

Misconception spotting education / misconception

Name the specific misconception behind a wrong (or right) learner answer, so the tutor can target it.

JSON {"question", "answer"} - a learner's answer to diagnose. Plain text is skipped.

Hint gate education / hint_gate

Decide whether a stuck learner should get a hint, be left to try a little longer, or be handed to a human teacher.

Any text: a short log of what the learner just did or said while working on a task.

Message mood education / message_mood

Read the emotional tone of a learner's message and the topic it is about, to route encouragement and help.

Any text: a short message a learner sent during a lesson.

Games & real time games-realtime

For game studios and live-service teams: tiny decisions that must be answered in milliseconds: cheat suspicion, player intent, difficulty and board moves.

Cheat suspicion games-realtime / cheat_suspicion

Rate one match summary as clean, worth a human review or an automatic ban.

JSON match summary {player, hours_played, headshot_pct, reaction_ms, kd, reports, account_age_days}.

NPC player intent games-realtime / npc_intent

Work out what a player wants from an NPC from one free-text line, in English, Slovenian or German.

One line of free text the player typed to an NPC, in any language.

Difficulty adjustment games-realtime / difficulty_adjust

Decide per player whether the next level should be easier, the same or harder; act alone only when confident.

JSON player stats {deaths_last_10, clears_last_10, avg_level_time_s, quit_rate, streak} for the last 10 levels; streak is consecutive clears (positive) or deaths (negative).

Player support triage games-realtime / player_support

Route a player's message to the right support queue: refund, bug, cheat report, account recovery or feedback.

A player's message to support as free text, in English, Slovenian or German.

Industrial faults industrial-faults

For maintenance and plant teams: diagnose machine faults from telemetry, prioritise maintenance, validate sensors and sort alarm floods.

Fault diagnosis industrial-faults / fault_diagnosis

Read one telemetry snapshot of a compressor, pump, motor or drive and name the most likely fault.

JSON snapshot {"asset", "vibration_mm_s" (RMS velocity), "temp_c", "current_a", "pressure_bar", "rpm", "note"}. Plain text is skipped.

Maintenance priority industrial-faults / maintenance_priority

Decide whether a maintenance finding needs action now, a slot in the next planned window, or only continued monitoring.

A short maintenance finding as text, including the asset and its criticality (critical, important, standard).

Sensor validity industrial-faults / sensor_validity

Check a short reading series and tell a healthy sensor from a stuck, spiking or drifting one.

JSON {"sensor", "last_readings": [numbers, oldest first], "unit", optional "reference" and "context"}. Plain text is skipped.

Alarm flood industrial-faults / alarm_flood

In a burst of alarms, classify one marked alarm as root cause, consequence, duplicate or nuisance.

Text listing a burst of alarms with timestamps and tags, with one alarm marked as the one to classify (for example 'CLASSIFY: A3').

Market intelligence market-intel

Analysts and trading desks triage crypto, equity and commodity news: catalyst strength, source status, tags and thesis-relevant headlines.

Catalyst strength market-intel / catalyst_strength

Score how strongly a headline is likely to move the named asset's price, from noise to a major catalyst.

Any text: a market headline or short news item, ideally naming the asset or sector it concerns.

Source status market-intel / source_status

Classify how firm a market claim is: unverified rumour, confirmed by a credible source, or an official announcement by the party itself.

Any text: a market news item or social post making a claim about a company, token or asset.

Asset and direction tagging market-intel / asset_tagging

Tag a headline with the asset class it concerns, the sector, and the likely price direction for that asset.

Any text: a market headline or short news item.

Thesis reranking market-intel / thesis_reranking

Score ten headlines against three investment theses and surface the ones that actually bear on each thesis.

An investment thesis as text. It is scored against the 10 headlines in the corpus.

Public inbox public-inbox

Citizen e-mails to a municipality (SL, EN): route to the right department, recognise the request type, set the response deadline class and flag personal data.

Department routing public-inbox / department

Route each citizen message to the municipal department that owns the matter.

Any text: a message sent to the municipality's general inbox, in Slovenian or English.

Request type public-inbox / request_type

Register the legal type of each request automatically when confident; a clerk checks the unclear ones.

Any text: a citizen message (SL or EN). The confidence of the request type picks act (register automatically) / confirm (clerk confirms) / human (clerk classifies).

Response deadline public-inbox / deadline_class

Set the response deadline class of each message, from no legal deadline to 8 days or less.

Any text: a citizen message (SL or EN). The class follows a generic municipal policy stated in the question instructions.

Personal data flags public-inbox / personal_data

Flag personal data, special-category data and data about minors in each message before it is filed or published.

Any text: a citizen message (SL or EN) as it would be filed in the document system.

Research & documents research-docs

For analysts, knowledge teams and data teams: fact-check claims against evidence, grade answers with a rubric, tag notes and route questions safely to a data layer.

Claim check research-docs / claim_check

Check whether a piece of evidence supports, contradicts or does not settle a claim.

JSON {"claim", "evidence"} - the evidence text is compared with the claim. Plain text is skipped.

Answer vs rubric research-docs / answer_rubric

Grade an answer 0-4 against the rubric of its question, for exams, onboarding quizzes and eval sets.

JSON {"question", "answer", "rubric"} - the answer is graded against the rubric only. Plain text is skipped.

Note tagging research-docs / note_tagging

Tag a note with its topic, whether it needs follow-up and how sensitive it is.

Any text: one personal or team note, meeting memo or snippet.

Query routing research-docs / query_routing

Decide whether a question to the analytics layer can be answered, needs a clarification or must be refused.

Any text: one natural-language question sent to the company analytics layer.

Security operations security-ops

For SOC analysts and MSSPs: give alerts a verdict and severity, gate response actions, triage reported e-mails and label log lines.

Alert verdict security-ops / alert_verdict

Close the loop on a SIEM/EDR alert: true positive, benign true positive or false positive, plus a 0-4 severity.

An alert summary as text: rule name, entities, what happened and the context the analyst gathered.

Response action gate security-ops / response_gate

Choose the containment step for a triaged incident: isolate the host, reset credentials, monitor or close. Act alone only when confident.

Incident summary as text: what was detected, on which host or account, and what the analyst already confirmed.

Reported e-mail triage security-ops / phishing_triage

Employees press 'Report phishing': sort each report into phishing, spam, legitimate mail or the company's own awareness simulation.

A reported e-mail as text: From, Subject, key headers (SPF/DKIM, X- headers), body excerpt and link targets. Any of EN, SL, DE, HR.

Log line labeling security-ops / log_labeling

Label raw log lines in bulk: ATT&CK-style tactic, asset type, and whether an analyst ticket is needed.

One raw log line (syslog, Windows event, firewall, web server, cloud or identity-provider audit).

Trading compliance trading-compliance

Market-conduct surveillance for energy and financial trading desks: inside-information checks, suspicious order patterns, trader chat review and regulatory report routing.

Inside-information check trading-compliance / inside_information

Decide whether an internal message about assets, outages or capacity must be disclosed before trading, is not inside information, or needs legal review.

Any text: an internal message, operations log line or e-mail about the company's assets, capacity or market (EN or SL).

Order pattern surveillance trading-compliance / order_patterns

Classify an order-book summary as layering/spoofing, wash trading, marking the close or normal activity.

JSON summary of one account's orders in a window: instrument, window, orders (counts, lifetimes, fills) and optional price_move, same_beneficial_owner, context. Plain text is skipped.

Trader chat surveillance trading-compliance / chat_surveillance

Screen trader chat lines: escalate clear misconduct, queue ambiguous lines for review, clear normal business.

Any text: one trader chat or e-mail line (EN or SL). The confidence of the action picks act / confirm / human.

Regulatory report routing trading-compliance / report_routing

Label each compliance event with the report it belongs to and its asset class.

Any text: a short description of a trade, order set, data feed or incident that may need regulatory reporting.

Images & video vision

Decisions from a photo, a screenshot, a chart or a short clip (frames): document intake, photo quality, screenshot triage, chart reading, motion and status lights.

Document photo intake vision / document_photo

A photo of a receipt, invoice, shipping label or handwritten note: what is it, can it be read, is the total above 100?

One photo or scan of a document.

Product photo quality gate vision / photo_quality

Marketplace or catalog upload check: can this product photo be published as is, or does it need a retake?

One product photo.

Screenshot triage vision / screenshot_triage

A user sends a screenshot of a message box: how serious is it, and does a person need to look?

A screenshot of an application message or alert dialog.

Chart trend reading vision / chart_trend

A dashboard screenshot or chart image goes in; the trend of the line comes out. Useful for reports and monitoring snapshots.

A line chart image.

Bar chart leader vision / bar_leader

Which category leads in a bar chart image, A to D.

A bar chart image with four labelled bars.

Video: motion direction vision / video_motion

Six frames of a short clip: which way does the object move, and does it cross the line? The client extracts the frames.

A short clip as 2-8 frames in time order.

Video: status light vision / video_status_light

Camera on a machine status light: which lamp is lit at the end of the clip and did it change? The client extracts the frames.

A short clip as 2-8 frames in time order, showing a three-lamp status light.

Working time & absence workforce-time

For HR, payroll and shift planners: classify absence messages, check timesheets against EU-style rest, break, night and overtime rules, and route corrections and shift swaps.

Absence messages workforce-time / absence_type

Label free-text absence messages from employees (EN, SL, DE, HR) with absence type, timing and language.

A short message from an employee to their manager or HR (chat, e-mail, SMS), in any of EN, SL, DE, HR.

Timesheet rule check workforce-time / timesheet_check

Check one day record against daily rest, weekly rest, night shift length, breaks and overtime caps, and name the rule that is breached.

JSON day record {"employee", "prev_shift_end", "start", "end", "break_min", "days_in_row", "week_overtime_h", "year_overtime_h"} with ISO local times. Overtime totals and days_in_row already include this day. Plain text is skipped.

Correction requests workforce-time / correction_gate

Route clock-in/out and absence corrections: approve automatically, ask the manager, or send to HR, with a confidence gate on top.

JSON correction request {"requested_by", "relation" (self | direct_manager | other), "record_date", "submitted", "period_locked", "kind", "change", "minutes_delta", "new_overtime_h", "reason"}, optional "rest_after_change_h". Plain text is skipped.

Shift-swap validation workforce-time / shift_swap

Validate a shift swap between two employees: approve, reject for rest, qualification or hours, or send it to the manager.

JSON swap request {"shift": {start, end, unit, requires}, "giver", "taker": {name, unit, qualifications, prev_end, next_start, week_hours_after, week_overtime_after_h}, "submitted"}. All times ISO local. Plain text is skipped.