Adds a full login/OTP/search/capture flow for United Ucard (OneHealthcareID
-> UHC Dental provider portal), mirroring the United SCO/DDMA pattern:
persistent Chrome profile, OTP polling, member eligibility search, and PDF
capture saved to the patient's Eligibility folder.
Also fixes a validation bug in the browser-safe patient schema where
email/address/city/zipCode/gender/phone were optional but not nullable,
rejecting patient records with null values on those fields during edits
(e.g. "Expected string, received null").
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
- Auto-fill service line price when a recognized CDT code is typed
(e.g. D2740), instead of requiring a separate "Map Price" click
- Fix chatbot preauth prefill using an empty DOB for price lookup,
causing age-tiered codes (D2740) to price as $0
- Re-apply fee-schedule pricing with the loaded patient DOB before
MassHealth preauth auto-submit
- Add a "Recement Crown" (D2920) combo button distinct from "Crown"
(D2740) to avoid picking the wrong code/price
- Fix AI note parsing to prioritize "recement" over "crown" regardless
of word order or tooth-number format (e.g. "#7 crown recement")
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
closeClaim() was clearing chatbotAutoSubmitSiteKey synchronously right
after the selenium job kicked off, before the job's async completion
handler could check it. This was masked for all but the last claim in
the batch because closeClaim() also navigated away (unmounting the
page) whenever more claims were queued. On the last claim the page
stayed mounted, so the completion handler read a stale cleared value
and let the PDF preview popup through. Track the site key in a ref
that survives closeClaim() and is cleared only after the job's own
completion handler consumes it.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
- 2PA with no explicit tooth numbers now defaults to D0220 #8 / D0230 #24
- Claims submitted via the AI chat column no longer auto-open the PDF preview modal
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
- unitedSCOEligibilityProcessor: overwrite patient's stored insuranceId when
the eligibility page reports a different (updated) Member ID than the one
searched with.
- CMSP worker: add AngularJS late-binding retry to results-table extraction
(ported from the MH worker's a0417653 fix), since every patient now routes
through this worker for the accumulator check.
- cmspEligibilityHistoryRemainingProcessor: stop forcing patient status to
UNKNOWN when extraction fails; a failed run landing after a successful one
was clobbering a correct ACTIVE/INACTIVE status.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Frontend uses the Web Speech API for speech-to-text and text-to-speech
(no key needed). Adds POST /api/translate backed by Google Cloud
Translation API v2, gated on GOOGLE_TRANSLATE_API_KEY (placeholder
added to .env/.env.example, unset by default).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Google Translate-style side-by-side layout with per-side language
pickers on the patient connection page. UI shell only for now, no
translation provider wired up yet.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Scripts still referenced the old project folder name after the rename,
which broke the desktop shortcuts and start/stop scripts.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Twilio now handles only the in-browser dial pad (Access Token + browser-Device
outbound TwiML); Telnyx takes over SMS, inbound voicemail, the AI voice
conversation loop, and staff-triggered outbound calls via TeXML. Adds
TelnyxSettings model, Communication.provider discriminator, new comms/routes
modules, and a real Telnyx Settings page.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
- Add remainingBenefit and thirdPartyInsurance columns to Patient; extract both
from the MassHealth accumulator/member-detail pages during the eligibility
check and surface them in the eligibility table
- Add an updated-date display and a toggle to suppress the post-check PDF
viewer popup (default off) on the insurance eligibility page
- Rename CMSP-prefixed PDF filenames and UI copy to MH, since CMSP and MH now
share the same worker
- Fold CMSP into MassHealth/MH across patient/claim insurance-type dropdowns
and the chatbot's eligibility routing, since they're the same insurance
- Remove auto-download of saved eligibility PDFs; downloading is now manual
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
DDMA now often skips the family-members results table and navigates
straight to the single-patient detail page, and labels the patient's
name "Subscriber name:" instead of "Name:". Update step1/step2 to
detect the detail-page URL correctly (it's nested under the search
path, so a plain substring check couldn't tell them apart), extract
the name from the new label, and fall back to parsing "Coverage is:"
for eligibility status when there's no results row to read it from.
Also fix a status-text substring bug ("active" matching inside
"inactive") and stop the backend from collapsing an unknown
eligibility result into INACTIVE. Skip name extraction entirely when
the caller already knows the patient (existing-patient recheck).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Only block appointment creation on explicit INACTIVE status, since
scrapes for some plan types (e.g. MassHealth Child Medicaid) can land
on UNKNOWN without actually finding an ineligible record.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Adds Levenshtein-based fuzzy word matching to lookupCdtCodes so
misspellings like "comprehensiv exam" resolve the same as
"comprehensive exam". Both the AI chat bot and the AI claim column
share this matcher, so the fix applies to both.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
D-Bus service activation can respawn gnome-keyring-daemon even after
the package is removed and PAM/autostart hooks are disabled.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
localhost resolves to ::1 first; backend/Vite only bind IPv4, causing
intermittent 502s when nginx tried the IPv6 loopback address.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
DentalHub's upload input is a multi-file input; chromedriver send_keys
appends to the existing selection instead of replacing it, so the second
iteration re-uploaded the first file (2 attachments became 3). Clear the
input value before each send_keys and skip duplicate filePaths.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Voice-call bookings previously only spoke the confirmation during the
live call, leaving nothing for the patient afterward. Now every
successful booking (recall, reschedule, and new-patient) also sends a
standalone confirmation text, regardless of which channel the
conversation happened on.
When the AI rejects a requested time in the recall booking flow, it reused
the same "Great, what time..." template used for a fresh ask, producing an
awkward double-greeting right after a rejection, and appended office hours
as a bare fragment instead of framing them as a sentence.
- Reset conversation stage to a fresh start on every new voice call
(recall and reminder alike), and on every hangup/timeout/goodbye,
so an abandoned or completed call never leaves a patient stuck
mid-flow for their next call or text.
- Add a staleness check to the SMS flow: after a long idle gap,
check whether the patient's reply still reads as a continuation
before resuming the stored stage, falling back to a friendly
"it's been a while" re-greeting when unsure.
- Fix parseTime() to handle Twilio's speech-to-text output, which
renders "pm"/"am" with periods and often drops the colon in a
spoken time (e.g. "100 p.m." -> 1:00 PM) — this was silently
failing and causing the AI to loop asking for a time it had
already been given.
- Extend the voice Gather timeout so the AI doesn't hang up on a
patient who just needs a few seconds to respond.
- Reword the recall-call opening line and rework the post-"yes"
prompt to ask for day and time together.
- Fix rclone WebDAV server to serve only the backups folder instead
of the entire backend app directory (was exposing .env and source).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
packages/db/usedSchemas/browser.ts (the browser-resolved copy of the Patient
zod schema, kept separate from the Node copy since it can't import
Prisma-generated code) was missing several fields added to the real schema
over time — preferredLanguage, gender, insuranceProvider, insuranceId,
groupNumber, policyHolder, and the PLAN_NOT_ACCEPTED status. Since the schema
isn't strict, those fields were silently stripped from every patient
create/update in the browser with no validation error, so edits to them
never reached the server.
Also fixes the AI call greeting always playing in English regardless of the
patient's preferred language — the greeting text was a single hardcoded
string, unlike every later turn in the conversation which already branched
by language.
Adds a new recall conversation flow (recall-graph.ts) triggered from the
Patient Connection "AI Call" button: offers an existing patient a routine
exam & cleaning, classifies their yes/no reply, then books directly onto
the schedule (Column A, 30-minute slots, real conflict checking against
office hours and existing appointments) instead of only capturing a
preference for staff to follow up on.
Also adds a "Voice call" label in the message thread to distinguish call
transcripts from SMS, and updates the Call Templates settings UI with an
editable recall greeting.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
AI phone calls now route through the same routePatientTurn() rule engine
as SMS (office-hours checks, slot-conflict checks, MassHealth eligibility,
staff escalation) instead of a freeform LLM chat, and speak in the
patient's preferred language (English/Spanish). Adds a separate "Call
Templates" settings section so spoken wording can be customized
independently from SMS wording.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Adds an "AI Call" option alongside Call/SMS/Chat that places an outbound
call where Lisa converses live with the patient via Twilio speech
gather + TTS, using the same conversational AI as chat. Includes a
separate, user-editable Call Template (Settings > AI Chat/Call Settings)
so the call greeting can be customized independently of SMS templates.
Also fixes the outbound-call webhook URL to always use the fixed public
Twilio hostname (CLOUDFLARE_HOST) instead of deriving it from the
triggering request, since staff-browser requests arrive over the
LAN-only hostname, which Twilio can never reach.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
The three Python services (PatientDataExtractorService, PaymentOCRService,
SeleniumService) provisioned their venvs via postinstall, chained into
root's own postinstall. A failure in any one of them aborted the whole
npm install, which could leave unrelated Node deps (e.g. Backend's sharp)
uninstalled — silently breaking the app with no indication why.
Renamed each service's postinstall to an explicit "setup" script, added a
root "npm run setup:python" to run all three, and documented it as its
own step in the README instead of an automatic install-time side effect.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Fresh-machine setup guide didn't cover the Payment OCR Service's Google
Vision credentials — add it as its own section (enable API, create
service-account key, install as google_credentials.json, verify).
- README: step-by-step Google Cloud Vision API key creation and setup
(enable API, create service-account key, place as google_credentials.json)
- ocrLocate: split into getOcrWords()/matchWordsForText() so a screenshot is
OCR'd once and matched locally against multiple text targets, instead of a
separate billed Vision API call per target (patient-row and Exam steps each
cut from 2-3 calls down to 1)
Claude vision's self-reported text coordinates had a small but consistent bias,
causing the patient-row double-click to land above the name and the Exam-column
divider count to occasionally miscount and hard-fail. Swap patient-row, Exam,
and Save button targeting to exact OCR bounding boxes (Google Vision via the
existing PaymentOCRService) instead of AI-estimated ratios.
schema.prisma had autoMhCheckEnabled/DayOfWeek/Hour on User (added in
fc3e8c0e) with no corresponding migration, so migrate deploy on a fresh
DB left those columns missing and broke seeding. Also drops
lab_rx_template.doctorName, a column that existed only in the DB (added
by an old migration) but was never in schema.prisma or referenced by
any app code.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Clicking "Add a file" triggers input.click() which opens the OS-native
file picker that WebDriver can't control, causing send_keys on the
underlying input to fail with a message-less exception. The input is
already present in the DOM (display:none) so send_keys can target it
directly without the click.
_fill_combobox typed the value then used a single unioned XPath
(match-by-text | option[1]) to pick the dropdown option. Selenium's
find_element on a multi-match XPath returns the first node in
document order, not the first union alternative, so option[1] (e.g.
tooth "1") was always clicked regardless of what was typed (e.g. "22").
Split into a value-match wait with a separate fallback wait for
option[1] only when no match is found.
Deleting/replacing an appointment attachment left an orphaned ClaimFile
copy behind (auto-linked earlier by autoLinkAppointmentFiles), causing
the Selenium claim automation to upload extra files beyond what's
currently attached to the appointment. Claim PUT was also append-only
for claimFiles, so removing an attachment in the UI never persisted.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Row selection previously computed the target row's y arithmetically
(header.y + rowHeight * rowIndex), which drifted further off-row the
deeper the target patient was in the filtered list, and an AI
"onTarget" verification step added to catch drift instead confirmed a
click point that was visibly in blank space below the grid.
locateAllOnScreenshot now reports each text match's top/bottom edges
instead of a self-estimated center, and the patient row step locates
the target's first and last name directly, pairing whichever
occurrences land on the same row instead of trusting arithmetic or a
qualitative AI judgment call.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Replaces fragile whole-screen text matching with a screenshot-diff step that
finds each dialog's actual pixel bounds, then restricts every subsequent AI
locate/click to that cropped region — eliminating false matches from text
elsewhere on screen (title bars, side panels). Adds column-boundary and
row-height detection so the patient row and Exam procedure click positions
are computed geometrically instead of relying on repeated fuzzy AI guesses
for visually similar neighbors. Also adds per-run screenshot/debug-crop
backups and a structured run.log for diagnosing failed runs, plus a
cmd:move primitive on the Windows agent for pre-click confirmation crops.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Adds a Windows-side agent (screenshot/click/type over HTTP) plus backend
services to locate UI elements via vision and drive an existing-patient
appointment flow in Open Dental, wired into the Copy/Type Agent page and
socket progress updates.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
- ccaPreAuthProcessor: keep claim status as PREAUTH after successful
submission instead of jumping straight to APPROVED, matching the
Tufts SCO/UnitedDH/MassHealth preauth flows; also tag PNG fallback
screenshots with image/png instead of the default application/pdf
so they render in the Documents preview modal.
- claims-recent-table: add a PREAUTH status case so it displays as
"Submitted" instead of falling through to the raw enum value.
- Fix NameError (col_idx vs col_expr) that silently corrupted grid-fill logs
- Poll for new grid rows to render before clicking, instead of failing immediately
- Abort submission with a clear error if code/tooth/billed amount fail to fill,
instead of silently submitting an incomplete claim
- Compute grid column positions live from header metadata instead of stale
hardcoded indices; billed amount was landing on a hidden helper column
(TAB_FROM_CODE_CELL) via td[last()] instead of the real Billed Amt column
- Read claim/encounter number from the top (newest) dashboard row instead
of the bottom row
skipAppointmentAttachmentMerge was blanket-set for any autoSubmit path
(chatbot, Claim with AI batch automation, insurance-status auto-check),
so screenshots/files attached directly on the appointment were silently
dropped from the MassHealth Selenium upload whenever a claim was
auto-submitted. Now the merge is only skipped if the automated flow
actually supplied its own files.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
- Chatbot: submitting with only a screenshot attached (no text) now triggers the same
"AI Detect & Check Eligibility" flow as the Copy Agent page
- detect-eligibility-info now also extracts the visible insurance payer name and picks
Tufts SCO auto-check when both Sun Life and DentaQuest are detected, else MassHealth
- Chatbot-attached files are now handed off to both claim and preauth forms unconditionally,
not just claims
- TuftsSCO preauth Selenium worker: verify-and-retry the Tooth field (typing was silently
getting reset by duplicate-procedure-code warning banners), add a final re-verification
pass across all rows, fix acknowledgement-checkbox targeting/verification, and verify
the "Next step" click actually advances the wizard instead of trusting a blind click
- selenium-claim route no longer re-appends previously saved appointment
attachments when the chatbot's autoSubmit flow already sent every intended
file, preventing an unwanted extra upload (e.g. attach 2 -> Selenium sees 3).
- Add a "Take screenshot" button to the AI chatbot's attachment bar, reusing
the same capture/crop flow as the appointment editor, so claims and
preauths can be submitted from the chatbot with a live screenshot attached.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
- Add "surg ext" alias mapping to D7210 in the AI CDT lookup.
- Add GET/DELETE endpoints for appointment attachments, and load existing
attachments in the appointment edit modal (they were uploaded correctly
but never re-fetched, so they appeared to vanish on reopen).
- Fix claim-form.tsx so previously-saved appointment attachments are always
loaded (even when a claim already exists) instead of only when creating a
brand-new claim, preventing them from being silently wiped on save.
- Merge existing appointment attachments into every insurance submit handler
(MassHealth, CCA, DDMA, UnitedDH, TuftsSCO) so Selenium submission always
includes files saved earlier via the Schedule editor, not just files
uploaded in the current claim-form session.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Adds an optional Provider (NPI) select to the Edit Appointment form,
sourced from Settings > NPI Providers. The AI "Claim for Column" batch
flow now prefers the appointment's own provider when resolving which
provider Selenium should select on the insurance site (e.g. MassHealth
rendering provider dropdown), falling back to the existing per-procedure
selection, active claim, or first configured provider.
Also resyncs packages/db's committed compiled schema/client artifacts
with their .ts sources — they had drifted out of sync (some untouched
since July), which was silently stripping the new npiProviderId field
via stale .strict() zod schemas.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
- Member ID and DOB in the AI chat "patient found" confirmation card are now
editable inputs, and edits are persisted back to the patient record before
proceeding so corrections don't revert to stale values.
- Fix auto-lookup-by-member-ID effect on the insurance status page clobbering
a hand-edited DOB with the stale DB value during chatbot prefill.
- CMSP (under-14) eligibility flow was missing the appointment-creation hook
that the adult MassHealth flow already had, so under-14 patients never got
scheduled after a chatbot-initiated "eligibility & appointment" request.
- For already-known patients, the chatbot was creating the appointment before
running the eligibility check instead of after; now eligibility is always
checked first and the appointment created only afterward, and only when the
resulting patient status is ACTIVE.
MassHealth eligibility lookups via the AI chatbot only matched existing
patients on memberId+DOB, so patients imported from PDF extraction (which
have a member ID but no stored birthday) were treated as new/unknown even
though they already exist in the system.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
GiteadandClaude Sonnet 5