sync: auto-sync from GURU-5070 at 2026-07-23 08:05:57

Author: Mike Swanson
Machine: GURU-5070
Timestamp: 2026-07-23 08:05:57

.claude/memory/MEMORY.md

@@ -244,3 +244,4 @@
 - [LAB-SVR is current](feedback_labsvr_retired.md) — Len's Auto LAB-SVR is the CURRENT server (replaced old LAB-SERVER); powered back on 2026-07-18
 - [Datto AV uses Bitdefender drivers](feedback_datto_av_bitdefender_drivers.md) — rtp1/rtp2/BdNet/BdSentry/netprotection_network_filter* are Datto AV's own drivers (Bitdefender SDK), NOT orphan Avira installs; check infocyte SDK dir before treating as orphans
 - [Prospect / call-list vetting criteria](feedback_prospect_list_vetting.md) — dedupe purchased lead lists against Syncro customers AND contacts by email domain; strip competitors, in-house-IT orgs, micro/consumer shops, HIPAA providers, non-AZ; location_type "Remote" = the contact, not the company
+- [Syncro hides disabled customers](feedback_syncro_disabled_customers_hidden.md) — GET /customers AND ?query= silently omit disabled (former) customers; 5082 vs 5375 with include_disabled=true; an empty search is NOT proof of absence

.claude/memory/feedback_prospect_list_vetting.md

@@ -1,48 +1,80 @@
----
-name: Prospect / call-list vetting criteria for ACG
-description: When vetting a purchased lead list for an outside appointment-setting firm, dedupe against Syncro (customers AND contacts, by email domain) and strip five categories Mike does not want onboarded.
-type: feedback
----
-
-Mike periodically buys a Tucson-area lead list and has it vetted before handing it to an
-outside sales / appointment-setting firm. Confirmed criteria (2026-07-22, on
-"Arizona Computer Guru New List.csv", 238 rows -> 168 callable / 70 excluded).
-
-**Why:** the setting firm calls on ACG's behalf, so a call into a current client, a
-competitor, or a prospect ACG would refuse to onboard is an embarrassment ACG pays for.
-
-**How to apply:**
-
-1. **Dedupe against Syncro, not the wiki.** `wiki/clients/` holds ~75 articles; Syncro holds
-   ~3000 customers + 768 contacts. Match on **email domain** (customer `.email` AND
-   `GET /contacts` — a lead is often a contact under a differently-named customer record),
-   then normalized business name, then fuzzy. Domain matching found every real hit; name
-   fuzzing at <0.86 produced only false positives. Exclude free-mail domains from the map.
-2. **Active vs former:** a Syncro record alone means nothing — many have zero tickets. Pull
-   ticket + invoice counts and max dates per matched customer. Recent invoices = ACTIVE
-   (exclude). Dormant = FORMER (keep on the list, flagged, as a warm re-approach).
-3. **Exclude these five categories:**
-   - **IT/MSP competitors** — MSPs, IT consultancies, dev shops, ISPs, web/marketing-tech firms.
-   - **Orgs with in-house IT** — the tell is the contact's own title (Sr Systems Engineer, Network
-     Administrator, IT Coordinator), plus universities, federal labs, districts, utilities,
-     hospitals, large law firms, public companies.
-   - **Micro / consumer-facing shops** — salons, spas, restaurants/bars, fitness studios,
-     single-chair practices, small retail. Low seat count, low IT spend.
-   - **Healthcare / HIPAA-heavy** — providers, clinics, hospices, home care, surgery centers,
-     diagnostic labs. NOTE: medical *device manufacturers* and *distributors* are fine — they
-     are not covered entities. Keep those.
-   - **Not Arizona** — verify HQ, do not trust the row's city/state.
-4. **`location_type` = "Remote" means the CONTACT is remote, not the company.** The row's
-   city/state is the contact's location. A Tucson company routinely shows a contact in another
-   state. Never drop for geography on that column alone — verify HQ (WebFetch the domain).
-   Flag them `remote_contact` so the setter knows onsite scheduling is harder.
-5. **Check whether the file actually contains contact data.** These vendor exports often ship
-   as previews: `work_email`, `work_phone`, `mobile`, `street1`, `company_revenue`,
-   `company_employee_count`, `primary_industry` etc. hold the literal strings `Yes`/`No`
-   (= "we have this field"), not values. Say so before Mike pays a setting firm to work a
-   list with no phone numbers in it.
-
-Output shape that worked: two CSVs in `~/Documents` — a tier-ranked `VETTED` call list
-(A = priority, B = fits, C = qualify first) and an `EXCLUDED` list with a reason per row so
-Mike can audit the cuts. See [[feedback-vault-every-credential]] for the unrelated-but-adjacent
-rule on anything credential-shaped that surfaces in these files.
+---
+name: Prospect / call-list vetting criteria for ACG
+description: When vetting a purchased lead list for an outside appointment-setting firm, dedupe against Syncro (customers AND contacts, by email domain) and strip five categories Mike does not want onboarded.
+type: feedback
+---
+
+Mike periodically buys a Tucson-area lead list and has it vetted before handing it to an
+outside sales / appointment-setting firm. Confirmed criteria (2026-07-22, on
+"Arizona Computer Guru New List.csv", 238 rows -> 168 callable / 70 excluded).
+
+**Why:** the setting firm calls on ACG's behalf, so a call into a current client, a
+competitor, or a prospect ACG would refuse to onboard is an embarrassment ACG pays for.
+
+**How to apply:**
+
+1. **Dedupe against Syncro, not the wiki.** `wiki/clients/` holds ~75 articles; Syncro holds
+   5375 customers + 768 contacts. Match on **email domain** (customer `.email` AND
+   `GET /contacts` — a lead is often a contact under a differently-named customer record),
+   then normalized business name, then **domain stem ignoring the TLD**, then fuzzy at >=0.86.
+   Exclude free-mail domains. All four passes are load-bearing:
+   - Open Enclose matched only on **name + domain stem** — Syncro has `openenclose.com`,
+     the list had `openenclose.net`.
+   - Peach Properties matched only on **domain** — filed in Syncro as "Good Rain Appartments".
+   - AZ Litho is in Syncro twice, as "AZ Litho" and "Arizona Lithography".
+   - Fuzzing below 0.86 gave only false positives (TMM Family Services -> AMP Facility Services;
+     NTH Consulting -> SRK Consulting; Stark Electric -> Fiesta Electric).
+
+   **`GET /customers` HIDES disabled customers — pass `include_disabled=true` or the entire
+   former-client population is invisible.** 5082 default vs 5375 with the flag = 293 hidden,
+   and those 293 ARE the former clients the task is about. `?query=` search hides them too and
+   returns empty, which reads exactly like "no such customer." Never conclude a company was
+   never a client from a default list or a query miss. See [[feedback_syncro_disabled_customers_hidden]].
+
+   **Compute page count from `meta.total_entries` and assert the pulled count matches.** A
+   hardcoded page guard silently capped a customer pull at 3000 of 5082 and the dedupe looked
+   complete while missing 41% of the tenant. `/contacts` also caps ~50/page regardless of
+   `per_page=100` — verify against `total_entries`, never trust the loop terminator alone.
+2. **Active vs former:** a Syncro record alone means nothing — many have zero tickets. Pull
+   ticket + invoice counts and max dates per matched customer. Recent invoices = ACTIVE
+   (exclude). Dormant = FORMER (keep on the list, flagged, as a warm re-approach).
+3. **Exclude these five categories:**
+   - **IT/MSP competitors** — MSPs, IT consultancies, dev shops, ISPs, web/marketing-tech firms.
+   - **Orgs with in-house IT** — the tell is the contact's own title (Sr Systems Engineer, Network
+     Administrator, IT Coordinator), plus universities, federal labs, districts, utilities,
+     hospitals, large law firms, public companies.
+   - **Micro / consumer-facing shops** — salons, spas, restaurants/bars, fitness studios,
+     single-chair practices, small retail. Low seat count, low IT spend.
+   - **Healthcare / HIPAA-heavy** — providers, clinics, hospices, home care, surgery centers,
+     diagnostic labs. NOTE: medical *device manufacturers* and *distributors* are fine — they
+     are not covered entities. Keep those.
+   - **Not Arizona** — verify HQ, do not trust the row's city/state.
+4. **`location_type` = "Remote" means the CONTACT is remote, not the company.** The row's
+   city/state is the contact's location. A Tucson company routinely shows a contact in another
+   state. Never drop for geography on that column alone — verify HQ (WebFetch the domain).
+   Flag them `remote_contact` so the setter knows onsite scheduling is harder.
+5. **The `Yes`/`No` columns are the vendor's data-availability manifest — this is BY DESIGN, not
+   a broken export.** `work_email`, `work_phone`, `mobile`, `alt_mobile`, `hq_phone`, `street1`,
+   `company_revenue`, `company_employee_count`, `primary_industry` etc. hold `Yes`/`No` meaning
+   "we hold this field for this contact." The vendor withholds the actual values until Mike
+   trims the list and sends it back; then they release the full records for the kept rows.
+   **Do not report this as a blocker.** Do mine it as a per-row reachability signal
+   (`phone_channels`) for the setter.
+
+**Deliverables — the send-back file is the one that matters.** Three CSVs in `~/Documents`:
+1. `<original name> - TRIMMED.csv` — **the vendor's own file: identical column set, column
+   order, and row order, kept rows only, rows byte-identical to source.** This is what goes back
+   to them to unlock the contact data. Verify col-list equality and row-order preservation
+   before handing it over.
+2. `ACG Call List - VETTED.csv` — tier-ranked working list for ACG (A = priority, B = fits,
+   C = qualify first) plus `setter_note`, `phone_channels`, `remote_contact`, `flag`, `notes`.
+   **Every FORMER-CLIENT row gets a `setter_note` telling the setter to handle them gently**
+   (Mike, 2026-07-23) — as the 2nd column so it can't be missed. Pitch depth to what the records
+   support: substantial multi-year relationship / real but limited work / on file with little or
+   no work. Overstating a relationship to someone who knows better is its own failure. Assert in
+   code that every former row has a note and no other row does.
+3. `ACG Call List - EXCLUDED.csv` — a reason per row so Mike can audit the cuts.
+
+Keep `matchId` and `domain` on the ACG-side files so the vendor's unlocked data joins back
+without redoing the vetting. See [[feedback-vault-every-credential]] for the adjacent rule on
+anything credential-shaped that surfaces in these files.

.claude/memory/feedback_syncro_disabled_customers_hidden.md

@@ -0,0 +1,47 @@
+---
+name: Syncro hides disabled customers from list AND search
+description: GET /customers and GET /customers?query= both silently omit disabled (former) customers. Pass include_disabled=true, and never treat an empty query result as proof a customer does not exist.
+type: reference
+---
+
+Verified live on the ACG tenant 2026-07-23:
+
+- `GET /customers?per_page=1` -> `meta.total_entries` = **5082**
+- `GET /customers?per_page=1&include_disabled=true` -> `meta.total_entries` = **5375**
+
+**293 customers are hidden by default**, and they are precisely the *former* clients — Syncro
+sets `"disabled": true` on the customer object when a relationship ends.
+
+**`GET /customers?query=...` hides them too, and returns an empty result set.** Customer 2549923
+has `business_name: "Open Enclose"`, `fullname: "Colt Barrins"`, `email:
+openenclose@openenclose.com`, 97 tickets and 122 invoices — and every one of these returned zero
+rows: `query=enclose`, `query=Open%20Enclose`, `query=Barrins`, `query=openenclose`. An empty
+search result looks identical to "no such customer" and is **not** proof of absence.
+`GET /customers/{id}` returns disabled customers fine — the id path is unfiltered.
+
+Probed 2026-07-23: only `include_disabled=true` changes the count. `disabled=true`,
+`archived=true`, `filter=disabled`, `status=disabled` are all ignored (still 5082).
+`GET /customers/autocomplete?query=` returned nothing useful.
+
+**How to apply:**
+
+- Any task about **former / lapsed / churned clients, win-back, or "have we ever worked with X"**
+  MUST pass `include_disabled=true`. Without it the answer is structurally wrong, not merely
+  incomplete.
+- **Never conclude "not a client" from a default list or a `?query=` miss.** If a human says a
+  company was a client and the search disagrees, the search is wrong — re-pull with
+  `include_disabled=true` and match on name + email-domain + domain-stem before doubting them.
+- Watch for **duplicate records** for one company, often one active and one disabled, sometimes
+  under different names (Open Enclose = 2549923 + 139892; AZ Litho = "AZ Litho" 21215131 disabled
+  + "Arizona Lithography" 2377491 active; SunDanzer = 139935 active + 35688623 disabled). Reconcile
+  by domain before reporting counts or history.
+- Email domains drift between the tenant record and the outside world (`openenclose.com` in Syncro
+  vs `openenclose.net` live), so match the **domain stem ignoring the TLD** as well as the full host.
+
+**Pagination, same session, same failure class:** derive page count from `meta.total_entries` and
+assert the pulled row count matches. A hardcoded `[ $P -gt 30 ] && break` guard capped a customer
+pull at 3000 of 5082 while the `N < 100` terminator never fired (every page was full), producing a
+confident 41%-incomplete result. `GET /contacts` separately caps at ~50/page regardless of
+`per_page=100` — 768 entries needs ~16 pages, not 8.
+
+Used by [[feedback_prospect_list_vetting]].

errorlog.md

@@ -19,6 +19,12 @@ Categories (the `[type]` tag): _(none)_ = skill/command execution failure ·
 
 <!-- Append entries below this line -->
 
+2026-07-23 | GURU-5070 | bash/api-pagination | [friction] Hardcoded page guard silently truncated a paginated pull: looped customers with '[ $P -gt 30 ] && break' at 100/page = capped at 3000 of 5082 real records, and the N<100 terminator never fired because every page was full. Produced a confident but 41%-incomplete dedupe. Always compute pages from meta.total_entries and assert the pulled count matches
+
+2026-07-23 | GURU-5070 | syncro/customers-search | GET /customers?query= returned ZERO results for 'enclose', 'Open Enclose', 'Barrins', 'openenclose' despite customer 2549923 having business_name='Open Enclose' / fullname='Colt Barrins'. Search silently excludes disabled customers - do not treat an empty query result as proof a customer does not exist [ctx: endpoint=/customers?query cust=2549923]
+
+2026-07-23 | GURU-5070 | syncro/customers-list | GET /customers omits disabled customers by default - 5082 returned vs 5375 with include_disabled=true (293 hidden). Former/inactive clients are invisible to the default list AND to ?query= search. Always pass include_disabled=true when the task is former-client detection [ctx: endpoint=/customers found=cust-2549923-OpenEnclose]
+
 2026-07-23 | GURU-BEAST-ROG | msp360 | request to https://api.mspbackups.com/api/Provider/Login failed: <urlopen error [Errno 11001] getaddrinfo failed> [ctx: command=monitoring] (x2)
 
 2026-07-22 | GURU-BEAST-ROG | rmm/hitmanpro | HitmanPro_x64 will not execute on Windows-on-ARM (Snapdragon X, LAPTOP-E5EKEJT8): process launches, idles at 0 CPU/1 thread, never scans, no log; 30-min RMM timeout hit. HMP has no ARM64 build - use Defender/EDR scan on ARM devices [ctx: client=bardach agent=d9cd756b cmd=e93474b9]

session-logs/2026-07/2026-07-23-mike-prospect-list-vetting.md

@@ -5,6 +5,12 @@
 - **Machine:** GURU-5070
 - **Role:** admin
 
+> **[WARNING] READ `## Update: 08:03 PT — Corrections` (end of file) BEFORE ACTING ON ANYTHING
+> BELOW.** Three findings in the original write-up are wrong: the Yes/No columns are NOT a broken
+> export (they are the vendor's data manifest, by design); the Syncro dedupe was incomplete
+> (`GET /customers` hides disabled customers AND the pull was truncated at 3000 of 5082); and the
+> former-client count is 16, not 9. The Update section is authoritative wherever the two conflict.
+
 ---
 
 ## Session Summary
@@ -217,3 +223,136 @@ CALL LIST: 168        EXCLUDED: 70
 - Tier legend: `A` = priority call (right size/industry, owner-exec contact, local);
   `B` = fits, lower priority; `C` = qualify first (geography, size, or unverified)
 - Exclusion reasons: `ACTIVE-CLIENT`, `COMPETITOR`, `IN-HOUSE-IT`, `MICRO`, `HEALTHCARE`, `NOT-AZ`
+
+---
+
+## Update: 08:03 PT — Corrections after Mike's feedback
+
+Three corrections to the record above. **Sections written before this Update contain errors that
+are corrected here; read this section as authoritative where the two conflict.**
+
+### CORRECTION 1 — the Yes/No columns are NOT a broken export (retracts the "blocker")
+
+The Session Summary and Pending Tasks above describe the `Yes`/`No` availability flags as a
+blocker requiring the vendor to release an unlocked file. **Wrong.** Mike: *"They have the
+phone/email addresses, they just don't show me those things. This list is what they sent me to
+trim."* The flags are the vendor's data-availability manifest, by design — they withhold values
+until the list is trimmed and returned, then release full records for the kept rows.
+
+Mined the manifest as a per-row reachability signal instead. All 238 rows have a work email, a
+mobile, and a LinkedIn; the differentiator is the rest:
+
+| Channels held | Rows (of 168 kept) |
+|---|---|
+| mobile + alt-mobile + main line | 83 |
+| mobile + main line | 46 |
+| mobile + alt-mobile | 14 |
+| mobile only | 17 |
+| includes a direct work line | 8 |
+
+Only 8 of 168 have a direct desk line; 17 have no main-line fallback. Added as a
+`phone_channels` column on the VETTED file.
+
+**New deliverable:** `C:\Users\guru\Documents\Arizona Computer Guru New List - TRIMMED.csv` —
+the vendor's own file, 168 kept rows, all 56 original columns in original order, original row
+order, rows byte-identical to source. This is the file that goes back to them. Verified with an
+explicit column-list equality + row-order + row-identity check. No ACG-internal columns added.
+
+### CORRECTION 2 — the Syncro dedupe was materially incomplete (two compounding bugs)
+
+Mike flagged Open Enclose as a previous client. It was not in my match set. Root cause was two
+independent failures, both mine:
+
+1. **`GET /customers` silently omits disabled customers.** Default `total_entries` = **5082**;
+   with `include_disabled=true` = **5375**. The 293 hidden records ARE the former-client
+   population this task exists to detect. `GET /customers?query=` hides them too and returns an
+   empty result — I ran `query=enclose`, `Open%20Enclose`, `Barrins`, `openenclose` and got zero
+   rows for a customer whose `business_name` is literally "Open Enclose". An empty query result
+   is NOT proof of absence.
+2. **My own pagination bug.** The customer pull used a hardcoded `[ $P -gt 30 ] && break` guard
+   at 100/page = capped at 3000 of 5082, and the real terminator (`N < 100`) never fired because
+   every page was full. The dedupe ran against 59% of the tenant and was reported as complete.
+
+A third trap on this specific record: Syncro holds `openenclose.com`, the lead list has
+`openenclose.net`, so domain matching would have missed it even with a complete pull. Added a
+**TLD-agnostic domain-stem pass** to the matcher.
+
+**Corrected pull:** 5375 customers (`include_disabled=true`, page count derived from
+`meta.total_entries`, pulled count asserted equal), 768 contacts. Four match passes: domain,
+domain-stem, normalized name, fuzzy >= 0.86.
+
+**Six additional former clients found. None active — so list membership did not change (still
+168 / 70); only flags and notes changed.** FORMER-CLIENT count 9 -> 16.
+
+| Lead | Company | Syncro | History |
+|---|---|---|---|
+| 175 | **Open Enclose** | 2549923 (DISABLED) | **97 tickets, 122 invoices thru 2020-11**; prepaid block expired 2025-08-19. Dupe 139892 disabled/no activity |
+| 39 | **Geomechanics Southwest** | 7088458 | **47 tickets, 139 invoices thru 2022-07** |
+| 75 | Peach Properties | 167909 | 10 tickets + 9 invoices, 2015 — filed as "Good Rain Appartments" |
+| 208 | AZ Litho | 2377491 | 4 tickets + 3 invoices thru 2021-01; dupe 21215131 "AZ Litho" disabled/no activity |
+| 20 | Finman & Associates | 3472144 | 1 ticket + 2 invoices, 2015 (filed under "Larry Finman") |
+| 53 | American Play Systems | 344305 | 2 tickets thru 2016-01, 2 invoices thru 2014-02 (filed under "Debbie Moore") |
+| 162 | Lightning Ridge Cellars | 7088554 | record only, zero activity |
+
+Also found: SunDanzer has a second disabled record (35688623, no activity); Marana Chamber has a
+second record (1021936, 1 ticket 2014-09).
+
+Each of the four match passes was load-bearing — Open Enclose matched only on name + domain-stem,
+Peach Properties only on domain (name is unrelated), AZ Litho exists under two different names.
+
+Logged to `errorlog.md`: `syncro/customers-list` (include_disabled gap),
+`syncro/customers-search` (silent empty query result), `bash/api-pagination` (`--friction`,
+hardcoded page guard).
+
+### CORRECTION 3 — gentle-handling notes for former clients (Mike's instruction)
+
+Mike: *"They can all stay on the list, but since we used them in the past a note that the setter
+should be gentle with them should be added."* Added a `setter_note` column as the **second
+column** of the VETTED file (next to tier, so it cannot be missed). Three levels, matched to what
+the Syncro records actually support — overstating a relationship to a prospect who knows better
+is its own failure mode:
+
+- **`established`** (Open Enclose, Geomechanics) — "Substantial multi-year relationship, ended
+  around YYYY. Do NOT pitch as a cold prospect... let them lead. No hard close."
+- **`served`** (Peach, Earth Savers, Termagon, AZ Litho, Saguaro Solar, Finman, Aviva, American
+  Play Systems) — "We did work for them (last worked YYYY). Acknowledge the prior relationship up
+  front rather than introducing ACG cold. Reconnect first, qualify second."
+- **`onfile`** (Serrato, Mission Management, Swaim, Marana Chamber, Ironwood Tree, Lightning
+  Ridge) — "On file with us from around YYYY, little or no work performed. Warm, not cold — don't
+  overstate the relationship."
+
+Peach Properties was demoted from `established` to `served`: 10 tickets/9 invoices sounds
+substantial but all fall inside 2015, so "multi-year" would have been false.
+
+`classify.py` now asserts every FORMER-CLIENT row has a `setter_note` and that no non-former row
+carries one, so a later reclassification cannot silently drop a note.
+
+### Final state
+
+```
+CALL LIST: 168            EXCLUDED: 70
+  Tier A: 88                HEALTHCARE     18
+  Tier B: 57                IN-HOUSE-IT    18
+  Tier C: 23                MICRO          13
+  former client:  16        NOT-AZ         10
+  remote contact: 48        COMPETITOR      7
+  unverified:      4        ACTIVE-CLIENT   4
+```
+
+Deliverables in `C:\Users\guru\Documents\`:
+1. `Arizona Computer Guru New List - TRIMMED.csv` — send-back file (vendor's format, 168 rows)
+2. `ACG Call List - VETTED.csv` — ACG working list (tier, setter_note, phone_channels,
+   remote_contact, flag, notes, src_row, matchId)
+3. `ACG Call List - EXCLUDED.csv` — 70 cuts with a reason each
+
+### Still open
+
+- **Mike has not reviewed the 168 names for pre-Syncro former clients.** The corrected dedupe
+  covers everything in Syncro including disabled records, but a client that predates Syncro or
+  was never entered remains undetectable by any query. An alphabetical scan list was printed for
+  him; no response yet.
+- **Medical device makers/distributors kept** (Wells Johnson #197, NeuroVRD #65, Medfix #29,
+  Avery Therapeutics #15) as non-covered-entities. Reversible if Mike wants all
+  healthcare-adjacent gone.
+- **4 UNVERIFIED** remain: Third Act/gvfan (#68), Pallet Recyclers (#115), REPICON (#151),
+  Southwest Solutions (#159). Sites blocked or unreachable.