diff --git a/frontend/src/app/help/page.tsx b/frontend/src/app/help/page.tsx index 4a473d2..7eafe90 100644 --- a/frontend/src/app/help/page.tsx +++ b/frontend/src/app/help/page.tsx @@ -81,19 +81,20 @@ export default function HelpPage() {

How it works

    -
  1. You create a job with campaign details, target locales, and a programme (Retail/Prime/Brand)
    2. -
  2. You upload a source xlsx file containing the English copy
    3. +
  3. You create a job with campaign details, target locales, programme (Retail/Prime/Brand), TM files, and AI model
    4. +
  4. You upload a source xlsx file with the English copy, plus optional supplementary files (briefs, glossaries, examples) — gated per-locale or sent globally
  5. You launch the job — each locale is processed in parallel (up to 4 at a time)
    6. -
  6. The AI pipeline runs a single consolidated agent per locale that handles TM matching, ranking, transcreation, and compliance in one pass
    7. +
  7. The AI pipeline runs a consolidated agent per locale that handles TM matching, ranking, transcreation, and compliance in one pass. Large briefs (>70 lines) are split into batches with prompt caching so the system prompt + TM data are paid for once and reused.
  8. You review the output with confidence tiers, backtranslations, and rationale
    9. -
  9. You approve, request revisions, or add comments on each line
    10. +
  10. You approve, request revisions, or add comments on each line — comments appear in the exported xlsx alongside the AI rationale
  11. You export the final output as a formatted xlsx (Tab 1: output table, Tab 2: linguistic summary)

Processing time and cost

A typical 53-line brief takes about 2-4 minutes per locale and costs approximately $0.30-0.50 per locale (depending on model). Multiple locales run - simultaneously, so a 12-locale job completes in about 4 minutes total. + simultaneously, so a 12-locale job completes in about 4 minutes total. Briefs of + 150+ lines automatically batch with prompt caching to keep costs flat.

@@ -135,7 +136,7 @@ export default function HelpPage() { TM Files - Select one or more TM channels to load. The campaign channel is auto-selected; add others (e.g. MASS as fallback) for cross-channel reference + Tick the TM channels to load. Channels are populated dynamically from the TM Registry, so any newly-uploaded TM appears automatically. Leave the list empty to send no TMs (saves tokens; useful when supplying a cut-down TM via supplementary files) Locales @@ -187,6 +188,23 @@ export default function HelpPage() { (for OOH, Display, etc.) and line breaks are preserved in the output.

+

Supplementary files (optional)

+

+ On the Upload step you can attach any number of supplementary files + (briefs, term lists, examples, prior creative). Each file gets a + per-file locale dropdown: +

+ +

+ Supported formats: .txt, .md, .json, .csv, .tsv, .docx. + Their text is inlined into the agent's context under a + “Supplementary Material” heading. Other formats (PDF, images) + are noted by filename only. +

+

Step 3: Launch

Review the summary, optionally add a context/override prompt with @@ -235,10 +253,23 @@ export default function HelpPage() { If you launch 12 locales, the first 4 start immediately and the remaining 8 queue up.

+

+ Auto-refresh: the dashboard and the per-job page poll + every 5 seconds while any job is active, so the status updates without + you needing to refresh the page. +

If a locale fails, you can re-run it individually without affecting other locales.

+ +

Inputs sent to the agent

+

+ On each job's detail page there's an Inputs sent to the agent card + showing the AI model used, which TM files were loaded, the supplementary + files attached, and the context override text. Useful for tracking and + re-prompting on script updates. +

{/* Reviewing */} @@ -291,7 +322,13 @@ export default function HelpPage() {
  • Needs Revision — flags the option for rework
  • Comment — add notes or specific feedback
    • -

    All feedback is saved to the database and visible to other team members.

    +

    + All feedback is saved to the database and visible to other team members. + Reviewer comments are appended to the rationale cell in the xlsx export{" "} + (prefixed Reviewer (Name): <comment>) so they ship downstream + with the rest of the work. The xlsx is regenerated on each download, so + late comments are always picked up. +

    {/* Locales */} @@ -400,9 +437,21 @@ export default function HelpPage() {

    - TM files are organized by channel (13 channels including Mass, Value, Onsite, Outbound, - UEFA, BDA, and more) and locale. You can select multiple TM channels per job — the system - loads all selected TM files into the agent's context for cross-channel reference. + TM files are organized by channel (Mass, Value, Onsite, Outbound, + UEFA, BDA, PrimeSpeed, etc.) and locale. You can select multiple TM + channels per job — the system loads all selected TM files into the agent's + context for cross-channel reference. The Channel dropdown is populated dynamically + from the registry, so any newly-uploaded channel appears automatically. +

    +

    + Replacing a TM: upload a TM via Admin > TM Files with the same + channel + locale as an existing one and the new file overwrites the old one in + place — the next job for that channel + locale uses the new content. +

    +

    + Adding a brand-new channel: the Channel field on the upload page + is a free-text input with autocomplete suggestions. Type any new name (e.g. + PrimeCBM) and it's registered alongside the existing channels.

    Admins can upload and manage these files in Admin > TM Files and{" "}