Baserow CRM 80/20 — Execution Plan

Created: 2026-03-24 Owner: Dmitri Zasage, Solanasis LLC Status: COMPLETE — All 4 phases executed and verified 2026-03-25 Validated: 2026-03-24 Execution started: 2026-03-25 Scope: Free self-hosted Baserow only — no Premium, no n8n, no Apollo, no guild sharing Supersedes: deep-plan-baserow-crm-transformation-2026-03-24.md (broader plan, kept for reference) Instance: baserow.solanasis.com (self-hosted, Docker, Cloudflare tunnel, Database ID: 54)

Validation Results (2026-03-24)

Assumptions Validated

#	Assumption	Evidence	Status
1	Self-hosted Baserow at baserow.solanasis.com, DB ID: 54	Confirmed in `migrate_to_baserow.py` load_env() and BASE_URL	✅
2	JWT auth for schema ops, DB token for row CRUD	Confirmed in BaserowClient: `_jwt_headers()` for create_table/field, `_token_headers()` for batch_create_rows	✅
3	httpx for HTTP client	`import httpx` + `httpx.Client()` confirmed in existing script	✅
4	.env has BASEROW_BASE_URL, DB_TOKEN, EMAIL, PASSWORD, DATABASE_ID	Confirmed from code references (not reading actual credentials)	✅
5	Fireflies YAML frontmatter has fireflies_id, date, participants, type	Confirmed from `2026-03-24 - Jonathan Dekle-summary.md`	✅
6	Existing —plan/—run flag pattern	Confirmed in migrate_to_baserow.py argparse	✅
7	run-fireflies-sync.py architecture	Confirmed: calls `claude -p`, logs to solanasis-docs/logs/, straightforward to extend	✅
8	Table row counts (Tag:51, Org:66, People:179, Meeting Notes:37)	From plan writing time — not re-verified via API. Non-blocking.	⚠️ Unverified
9	Formula syntax (isblank, field, date_diff)	Plan’s syntax notes section (lines 563-568) is correct	⚠️ See Issue #4
10	link_row field creation via API	Generic `create_field()` method forwards dict — needs `link_row_table_id` param	⚠️ Untested

Issues Found — Must Fix Before Execution

Issue 1: Transcript count overstated (LOW impact)

Plan says: “Currently 70+ meetings exist only as markdown files” (line 106) Reality: 19 summary files (12 meetings + 7 voice notes) + 20 transcripts. Total ~48 markdown files, but many are paired summary+transcript of the same meeting. Unique meetings: ~17-20. Impact: Backfill processes fewer files. Script logic unchanged. Fix: Updated count below in Section 5 (“The Data Bridge”).

Issue 2: Idempotency not addressed (MEDIUM impact)

Problem: schema-upgrade-v2.py needs to handle “table already exists” and “field already exists” gracefully. If run twice, could fail or create duplicate fields/tables. Fix: Script must check list_tables() before creating; check list_fields(table_id) before adding fields. Skip if exists, log the skip.

Issue 3: “Dmitri Sunshine” vs “Dmitri Zasage” in name matching (MEDIUM impact)

Problem: Fireflies records use “Dmitri Sunshine” as participant name (confirmed in frontmatter). People table and business contexts use “Dmitri Zasage”. The Phase 3 name exclusion filter says “first non-Dmitri name” — needs to explicitly handle both variants. Fix: Exclude list should be {"Dmitri Sunshine", "Dmitri Zasage", "Dmitri"} (case-insensitive first-name match as fallback).

Issue 4: Formula function naming inconsistency (MEDIUM impact)

Problem: The “Formula Functions Available” section (line 72) lists is_blank(). The syntax notes (line 565) correctly say isblank() (no underscore). Someone following the feature matrix would write broken formulas. Fix: Corrected below — isblank() is correct Baserow syntax.

Issue 5: Voice notes in separate folder (LOW impact)

Problem: Phase 3 sync script says scan meeting-transcripts/*-summary.md, but 7 voice note summaries live in voice-notes/*-summary.md. Fix: Backfill should scan both meeting-transcripts/ and voice-notes/ folders for *-summary.md files.

Issue 6: Phase 2 execution method unclear (LOW impact)

Problem: 31 views + 4 forms described in detail, but no script or API reference for creating them. Is this manual UI work or scripted? Recommendation: Script view creation in schema-upgrade-v2.py Phase 2 section. Baserow API supports POST /api/database/views/table/{table_id}/ for view creation with filters/sorts. Forms use POST /api/database/views/table/{table_id}/ with type: "form". This keeps everything repeatable and idempotent.

API Pattern Confirmed (from existing codebase)

The link_row field creation pattern is confirmed in solanasis-scripts/baserow/src/migrate/schema.ts:

# Python equivalent of the proven TS pattern:
payload = {"name": "Person", "type": "link_row", "link_row_table_id": target_table_id}
# POST /api/database/fields/table/{table_id}/ with JWT auth

Critical execution order (already documented in TS migration, must replicate in Python):

Create ALL tables first (need table IDs for link_row references)
Create non-link fields on each table
Create link_row fields LAST (targets must exist)

The TS migration also demonstrates idempotency: check fieldMap[fieldDef.baserowField] before creating (skip if exists).

Solution Robustness Check

Simple without over-engineering — 4 phases, clear progression, no premature optimization
Maximum reuse of existing code — BaserowClient pattern from migrate_to_baserow.py
Follows project conventions — Python first, .env credentials, —plan/—run flags
No hardcoding where config should be used — DB ID from .env, options in schema dicts
Security considerations — .env for credentials, no hardcoding, follows CLAUDE.md security rules
Error handling — Idempotency requirement added (Issue #2 fix). Rate limiting handled in existing pattern.

Wrong vs Right Fix Comparison

Wrong Approach	Right Approach	Why
Create tables/fields manually in UI	Script everything (idempotent)	Repeatable, survives mistakes, documents schema
Hardcode table IDs in scripts	Discover via list_tables() dynamically	IDs change between environments
Create link_row fields alongside other fields	Create ALL tables first, then link_row fields	link_row needs target table to already exist
Scan only meeting-transcripts/ for Fireflies	Scan both meeting-transcripts/ AND voice-notes/	7 voice note summaries would be missed
Filter Fireflies participants by “Dmitri Zasage”	Filter by {“Dmitri Sunshine”, “Dmitri Zasage”, “Dmitri”}	Fireflies uses “Dmitri Sunshine”
Build 70-file backfill expectations	Plan for ~19 summary files	Actual count is 19, not 70

Validation Verdict

VALIDATED — Ready for execution. All 6 issues have been corrected in-place. No blockers remain. The plan is sound, the API patterns are proven in the existing codebase, and the execution order is clear.

Decisions Made (2026-03-24)

Decision	Answer
Self-hosted migration	COMPLETE — working with self-hosted instance
Premium license	NO — free tier only
n8n automation	DEFERRED — separate project later
Apollo.io import	DEFERRED — separate project later
Guild CRM sharing	VISION ONLY — deferred indefinitely, do not design for it now
Community plugins	Only if free and high-value (80/20 rule)

What Free Self-Hosted Baserow Actually Gives Us

This is the definitive feature matrix from official docs — know this to avoid wasted effort:

Views (How You See Data)

View Type	Free?	CRM Use
Grid	YES	Primary working view. Group by field = poor man’s kanban. Filter + sort = dashboards.
Form	YES	Data entry for contacts, interactions, deals, tasks
Gallery	YES	Contact cards with photos/avatars
Kanban	PREMIUM	Not available — use grouped grids instead
Calendar	PREMIUM	Not available — use date-sorted grids instead
Dashboard	PREMIUM	Not available — use filtered grids + formula fields instead
Timeline	PREMIUM	Not available

Fields (The Real Power — ALL Free)

Field Type	Free?	CRM Use
link_row	YES	THE key feature — relations between tables
lookup	YES	Pull data from linked records (read-only)
rollup	YES	Aggregate linked data (sum, avg, count, etc.)
count	YES	Count linked rows
formula	YES	100+ functions — date math, conditionals, text, aggregations
single_select / multiple_select	YES	Status, stage, type dropdowns
All other types	YES	text, long_text, number, boolean, date, url, email, phone, file, rating, autonumber, created_on, last_modified, uuid, password, created_by, last_modified_by
AI prompt	PREMIUM	Not available — use external Python + AI API instead

Infrastructure (All Free on Self-Hosted)

Capability	Limit
Rows	Unlimited (server resources only)
Tables	Unlimited
Databases	Unlimited
Storage	Unlimited
API requests	Unlimited, no rate limit
Webhooks	Unlimited (row created/updated/deleted triggers)
Plugin support	Yes (requires custom Docker image build)

Formula Functions Available (All Free)

The formula engine is extremely powerful — 100+ functions including:

Text: concat(), upper(), lower(), replace(), search(), contains(), length(), split_part()
Math: round(), abs(), ceil(), floor(), mod(), power(), greatest(), least()
Date: now(), today(), day(), month(), date_diff(), datetime_format(), date_interval()
Logic: if(), and(), or(), not(), isblank(), when_empty()
Aggregation (across linked tables): sum(), avg(), min(), max(), count(), join(), filter(), any(), every()
Link/URL: link(), button(), get_link_url()
Special: field(), row_id(), lookup(), index()

Key insight: lookup() + aggregation functions in formulas = Airtable-equivalent rollups, for free.

The 80/20: What Actually Makes Free Baserow Amazing for CRM

These 5 things deliver 80% of the value:

1. Proper Relational Schema (The Foundation — 50% of Value)

Link_row fields + lookups + rollups turn Baserow from a spreadsheet into a relational database. Click a person → see their org, all meetings, all deals, all tasks. This is the #1 transformation.

2. Smart Grouped Grid Views (The Dashboard Replacement — 15% of Value)

Grid views grouped by a single_select field mimic kanban boards. Grid views filtered by date ranges mimic calendar views. Multiple filtered/sorted views per table = your daily operating dashboard.

3. Formula Fields Doing Heavy Lifting (The Automation Replacement — 10% of Value)

Formulas compute data that would otherwise require manual updates or external automation:

“Days since last interaction” = date_diff('day', lookup('Interactions', 'Date'), today())
“Deal count” = rollup counting linked deals
“Full display name” = concat(field('Name'), ' — ', lookup('Organization', 'Name'))

4. Forms for 60-Second Data Entry (The UX — 10% of Value)

4 forms bookmarked in browser toolbar = log an interaction in 60 seconds, add a contact in 30 seconds. This is what makes CRM discipline sustainable.

5. Fireflies → Baserow Sync (The Data Bridge — 15% of Value)

Meeting notes auto-flowing from Fireflies into the Interactions table (linked to People) closes the biggest data gap. Currently ~19 meeting/voice-note summaries exist only as markdown files (12 in meeting-transcripts/, 7 in voice-notes/).

Current State (What Exists in Baserow Today)

Tables (from Coda migration, now on self-hosted)

Table	Rows	Key Fields	Issues
Tag	51	Name, Notes	Fine as-is
Location	33	Name, State, Notes	Fine as-is
Organization	66	Name, Tags→Tag, Location→Location, Website, LinkedIn, Summary, Notes	Missing: Industry, Size Band, Account Status
People	179	Name, Tags→Tag, Location→Location, Title, Organization→Org, Phone, Email, LinkedIn, social URLs, Notes, Connected From, Referral Source	Missing: Status, Lead Score, Last Contacted, Next Touch Date
Meeting Notes	37	Person→People, LinkedIn, Date, Follow-up Date, Notes	To be migrated into new Interactions table

Pipeline Tables (flat, isolated — NOT linked to People/Org)

Table	Rows	Notes
Foundation Prospects	~200+	30 fields, tier scoring, financials
fCTO Partners	~50+	24 fields, tier scoring
MSP Prospects	~100+	Geographic + RIA density scoring

These stay as-is for now. Pipeline consolidation is deferred to the Apollo project.

Fireflies Sync

Runs hourly via Windows Task Scheduler → Claude Code
Saves markdown files to solanasis-docs/meeting-notes/meeting-transcripts/
19 summaries + 20 transcripts synced
Does NOT write to Baserow — this is the gap we’re closing

Execution Plan: 4 Phases

Phase 1: Schema Upgrade — New Tables + Enhanced Fields

What: Create new tables, add fields to existing tables, migrate Meeting Notes → Interactions Effort: ~2-3 hours in a single session Dependencies: None — do this first

1A. Create the Interactions Table

This is the most important new table. It replaces the existing Meeting Notes table and becomes the system of record for ALL touchpoints.

Field Name	Type	Config	Notes
Title	text (primary)	Required	e.g., “Call with Jonathan Dekle — 2026-03-24”
Person	link_row → People	Required	Who was this interaction with
Organization	link_row → Organization	Optional	Auto-populated from Person’s org, or manual
Date	date	Required, date_format: ISO	When it happened
Type	single_select	Options below	What kind of interaction
Channel	single_select	Options below	How it happened
Summary	long_text	Required	Key takeaways (2-3 sentences)
Outcome	single_select	Options below	How did it go
Next Step	text		What needs to happen next
Follow-up Date	date		When to follow up
Fireflies ID	text		For dedup on auto-synced meetings
Transcript Link	text		Relative path to markdown transcript
Source	single_select	Options below	How this record was created

Single select options:

Field	Options
Type	Call, Meeting, Email, LinkedIn Message, Referral, Voice Note, Conference, Webinar, Other
Channel	Zoom, Phone, In-Person, Email, LinkedIn, WhatsApp, Google Meet, Teams, Other
Outcome	Positive, Neutral, Negative, No Response, Rescheduled
Source	Manual, Fireflies, Email Sync, Import, Backfill

1B. Create the Deals Table

Field Name	Type	Config	Notes
Deal Name	text (primary)	Required	e.g., “Acme Corp — Security Assessment”
Organization	link_row → Organization		Company
Primary Contact	link_row → People		Main person
Service Line	single_select	Options below	What we’d deliver
Stage	single_select	Options below	Where in the pipeline
Estimated Value	number	number_decimal_places: 0, number_negative: false	Dollar amount
Probability	number	number_decimal_places: 0	0-100%
Expected Close	date		Target close date
Source	single_select	Options below	How we found them
Pipeline	single_select	Options below	Which vertical pipeline
Blockers	long_text		What’s preventing progress
Next Action	text		Immediate next step
Won/Lost Reason	long_text		Post-close learning
Created Date	date	date_format: ISO	Auto-set on creation
Last Activity	date		Updated by formula or manually

Single select options:

Field	Options
Service Line	Security Assessment, DR Verification, Data Migration, CRM Setup, Systems Integration, AI Implementation, fCIO Retainer, fCSIO Retainer, fCOO Retainer, Other
Stage	Lead, Discovery, Proposal Sent, Negotiation, Won, Lost, On Hold
Source	Referral, Cold Outreach, Inbound, Network Event, Conference, LinkedIn, Apollo, Fireflies Intro, Other
Pipeline	Primary, Foundation, fCTO, MSP, VC

1C. Create the Tasks Table

Field Name	Type	Config	Notes
Task	text (primary)	Required	What needs to be done
Person	link_row → People	Optional	Related contact
Organization	link_row → Organization	Optional	Related company
Deal	link_row → Deals	Optional	Related deal
Due Date	date		When it’s due
Status	single_select	To Do, In Progress, Waiting, Done	Current state
Priority	single_select	Urgent, High, Normal, Low	How important
Type	single_select	Follow-up, Proposal, Research, Admin, Outreach, Onboarding	What kind of task
Notes	long_text		Additional context

1D. Add Fields to People Table

New Field	Type	Config	Purpose
Status	single_select	New, Active, Warm, Cold, Client, Partner, Dormant, Do Not Contact	Contact lifecycle
Lead Score	number	0-100, decimal: 0	Manual score for now
Last Contacted	date		Date of most recent interaction
Next Touch Date	date		When to reach out next
Pipeline Source	single_select	Manual, Apollo, LinkedIn, Foundation, fCTO, MSP, VC, Fireflies, Calendly, Referral	Where this contact came from
Apollo ID	text		For future Apollo dedup
Interaction Count	formula or count	Count linked Interactions	Auto-compute from Interactions link
Deal Count	formula or count	Count linked Deals	Auto-compute from Deals link

Important: After creating the Interactions table with link_row → People, Baserow auto-creates a reverse link on People. Use that reverse link field for the count/rollup calculations.

1E. Add Fields to Organization Table

New Field	Type	Config	Purpose
Industry	single_select	Finance/Wealth Mgmt, Healthcare, Nonprofit/Foundation, Technology, Legal, Education, Government, Manufacturing, Consulting, Other	Vertical classification
Size Band	single_select	1-10, 11-50, 51-200, 201-1000, 1000+	Company size
Account Status	single_select	Prospect, Active Client, Former Client, Partner, Vendor, Do Not Contact	Relationship status
City	text		For geographic targeting
State	text		For geographic targeting

1F. Migrate Meeting Notes → Interactions

The existing 37 Meeting Notes rows need to move into the new Interactions table. Then archive (not delete) the Meeting Notes table.

Migration logic:

Read all 37 rows from Meeting Notes table
For each row:
- Read the Person link_row value → get the People row ID
- Look up that People row to get the Name
- Create an Interactions row:
  - Title = “Meeting with {Person Name} — {Date}” (or “Meeting — {Date}” if no person)
  - Person = same People row ID (link_row)
  - Date = Meeting Notes.Date
  - Type = “Meeting”
  - Summary = Meeting Notes.Notes (copy full text)
  - Follow-up Date = Meeting Notes.Follow-up Date
  - Source = “Backfill”
- If Meeting Notes.LinkedIn is set, search Organization table for matching LinkedIn URL → link Organization
Verify: 37 Interactions rows created, all Person links intact
Rename Meeting Notes table to “Meeting Notes (ARCHIVED — see Interactions)”

Script: solanasis-scripts/baserow/schema-upgrade-v2.py

Should use the existing Python patterns from foundation-pipeline/migrate_to_baserow.py:

JWT auth for schema operations (create table, create field)
DB token auth for row CRUD (read, create, batch)
httpx for HTTP client
.env for credentials (BASEROW_BASE_URL, BASEROW_DB_TOKEN, BASEROW_EMAIL, BASEROW_PASSWORD, BASEROW_DATABASE_ID)
--plan flag for dry run, --run for execution
MUST be idempotent: Check list_tables() before creating tables (skip if exists). Check list_fields(table_id) before adding fields (skip if exists). Log all skips. This allows safe re-runs after partial failures.

1G. Validation Checklist

Phase 2: Views & Forms — The Daily Operating Dashboard

What: Create filtered/grouped views for each table + 4 data entry forms Effort: ~1-2 hours (can be same session as Phase 1) Dependencies: Phase 1 complete Execution method: Script via Baserow API — POST /api/database/views/table/{table_id}/ supports grid, gallery, and form view creation with filters, sorts, and grouping. Include in schema-upgrade-v2.py as Phase 2 functions. Must be idempotent (check existing views by name before creating).

2A. People Table Views

View Name	Type	Configuration	Daily Use
All Contacts	Grid	Sort: Name A→Z	Master list, default view
Hot Leads	Grid	Filter: Status IN (Active, Warm). Sort: Lead Score DESC	Who to reach out to today
Needs Follow-Up	Grid	Filter: Next Touch Date ≤ today + 7 days, Status ≠ Do Not Contact. Sort: Next Touch Date ASC	Weekly follow-up review
Clients	Grid	Filter: Status = Client	Active client roster
Partners	Grid	Filter: Status = Partner	Referral network
Dormant	Grid	Filter: Last Contacted < (30 days ago) OR Last Contacted is empty, Status NOT IN (Client, Do Not Contact, Dormant). Sort: Last Contacted ASC	Re-engagement candidates
Recently Added	Grid	Sort: created_on DESC. Limit to page size 50	What’s new in the CRM
Contact Cards	Gallery	Card title: Name. Cover image: (none). Fields: Title, Organization (lookup), Email, Phone, Status, Lead Score	Visual contact browser

Pro tip: The “Hot Leads” view becomes your daily starting point. Open it each morning — it shows who to contact, sorted by score.

2B. Interactions Table Views

View Name	Type	Configuration	Daily Use
Recent Activity	Grid	Sort: Date DESC	Quick review — what happened lately
This Week	Grid	Filter: Date ≥ (start of current week). Sort: Date DESC	Weekly activity summary
Follow-ups Due	Grid	Filter: Follow-up Date ≤ today + 3 days, Follow-up Date is not empty. Sort: Follow-up Date ASC	Action items — don’t miss follow-ups
By Person	Grid	Group by: Person. Sort: Date DESC within groups	All touchpoints per contact
Meetings	Grid	Filter: Type = Meeting. Sort: Date DESC	Meeting history
From Fireflies	Grid	Filter: Source = Fireflies. Sort: Date DESC	Auto-synced entries

2C. Deals Table Views

View Name	Type	Configuration	Daily Use
Pipeline by Stage	Grid	Group by: Stage. Sort: Expected Close ASC within groups	This is your kanban replacement — grouped grid shows deals organized by stage
All Deals	Grid	Sort: Expected Close ASC	Full deal list
Active Deals	Grid	Filter: Stage NOT IN (Won, Lost). Sort: Expected Close ASC	What’s in play
Closing This Month	Grid	Filter: Expected Close within current month, Stage NOT IN (Won, Lost)	Focus on closing
Stale Deals	Grid	Filter: Last Activity < (14 days ago), Stage NOT IN (Won, Lost, On Hold)	Deals that need attention
Won	Grid	Filter: Stage = Won. Sort: Created Date DESC	Win tracking
Lost	Grid	Filter: Stage = Lost. Sort: Created Date DESC	Learn from losses

Pro tip: The “Pipeline by Stage” grouped view is the most important view in the whole CRM. When you group a grid by a single_select field, Baserow shows each group as a collapsible section with a count badge. It’s not drag-and-drop like kanban, but it gives you the same “deals organized by stage” view for free.

2D. Tasks Table Views

View Name	Type	Configuration	Daily Use
Today	Grid	Filter: Due Date = today, Status ≠ Done. Sort: Priority (Urgent first)	Morning task list
This Week	Grid	Filter: Due Date ≤ end of week, Status ≠ Done. Sort: Due Date ASC	Weekly planning
Overdue	Grid	Filter: Due Date < today, Status ≠ Done. Sort: Due Date ASC	Red flag view — fix these first
By Status	Grid	Group by: Status. Sort: Priority within groups	Visual task board (grouped = kanban-like)
By Priority	Grid	Group by: Priority. Sort: Due Date ASC within groups	Triage view
All Tasks	Grid	Sort: created_on DESC	Master task list

2E. Organization Table Views

View Name	Type	Configuration	Daily Use
All Organizations	Grid	Sort: Name A→Z	Master list
Active Clients	Grid	Filter: Account Status = Active Client	Client companies
Prospects	Grid	Filter: Account Status = Prospect	Sales targets
By Industry	Grid	Group by: Industry	Vertical analysis

2F. Forms (4 Bookmarkable Data Entry Points)

After creating forms, bookmark each URL in browser toolbar bar for 1-click access.

Form 1: Quick Contact Capture → People table

Visible Field	Required?	Default Value	Notes
Name	YES
Email	no
Phone Number	no
Title	no
Organization	no		Link_row dropdown of existing orgs
LinkedIn	no
Status	no	”New”	Pre-filled
Pipeline Source	no	”Manual”	Pre-filled
Notes	no

Hidden fields (not on form): Lead Score, Last Contacted, Next Touch Date, all social URLs, Apollo ID, Tags, Location, Connected From, Referral Source

Form 2: Log an Interaction → Interactions table

Visible Field	Required?	Default Value	Notes
Person	YES		Link_row dropdown
Date	YES	today	Pre-filled with today
Type	YES		Must select one
Channel	no
Summary	YES		2-3 sentences
Outcome	no
Next Step	no
Follow-up Date	no

Hidden fields: Title (auto-generate later), Organization, Fireflies ID, Transcript Link, Source (default: “Manual”)

This is the single most important form. If Dmitri uses only one form, this is it. Log every meaningful conversation in 60 seconds.

Form 3: New Deal → Deals table

Visible Field	Required?	Default Value	Notes
Deal Name	YES		e.g., “Acme — Security Assessment”
Organization	no		Link_row dropdown
Primary Contact	no		Link_row dropdown
Service Line	no
Stage	no	”Discovery”	Pre-filled
Estimated Value	no
Expected Close	no
Source	no
Pipeline	no	”Primary”	Pre-filled
Next Action	no

Form 4: Quick Task → Tasks table

Visible Field	Required?	Default Value	Notes
Task	YES		What needs to be done
Due Date	YES		When
Priority	no	”Normal”	Pre-filled
Person	no		Link_row
Deal	no		Link_row
Type	no
Notes	no

2G. Validation Checklist

Phase 3: Fireflies → Baserow Sync

What: Extend the existing Fireflies sync to also write Interactions rows in Baserow Effort: ~2-3 hours Dependencies: Phase 1 complete (Interactions table must exist)

3A. Current Fireflies Architecture

Windows Task Scheduler (every hour at :07)
  → solanasis-docs/scheduled-tasks/scripts/run-fireflies-sync.py
    → claude -p [sync prompt from 01-fireflies-sync-prompt.md]
      → Calls Fireflies MCP (get_transcripts, get_summary, get_transcript)
      → Saves markdown files to solanasis-docs/meeting-notes/meeting-transcripts/
      → Updates solanasis-docs/meeting-notes/fireflies-sync-status.md

The gap: Markdown files are saved but nothing touches Baserow.

3B. Target Architecture

Same pipeline as above, PLUS:
  → After saving markdown files:
    → Read participant names/emails from Fireflies data
    → Search Baserow People table for matching contact
    → Create Interactions row linked to People row
    → Update People.Last Contacted if this is more recent

3C. Implementation — Separate Python Script (Recommended)

After evaluating Options A (extend sync prompt) and B (separate script) from the broader plan, Option B is better for this 80/20 approach because:

It’s a standalone Python script that can be tested independently
It doesn’t add complexity to the Claude Code sync prompt
It runs after the markdown files are saved, using them as input
It follows the “Python first” scripting rule from CLAUDE.md

Script: solanasis-scripts/baserow/sync-fireflies-to-baserow.py

Logic:

1. Scan both `solanasis-docs/meeting-notes/meeting-transcripts/*-summary.md` AND `solanasis-docs/meeting-notes/voice-notes/*-summary.md`
2. For each summary file:
   a. Parse YAML frontmatter → get fireflies_id, date, participants, type
   b. Check Baserow Interactions table: does a row with this Fireflies ID exist?
      → If yes: skip (already synced)
   c. For the primary participant (first non-Dmitri name):
      - Search Baserow People table by name (exact match first)
      - If no match, search by email if available
      - If still no match: create a new People row (Name, Email, Status="New", Pipeline Source="Fireflies")
      - Get the People row ID
   d. Read the first 500 chars of the summary content as the interaction summary
   e. Create Interactions row:
      - Title: "{Type} with {Person Name} — {Date}"
      - Person: link to People row ID
      - Date: meeting date from frontmatter
      - Type: "Meeting" or "Voice Note" (from frontmatter type field)
      - Channel: "Zoom" (default, most meetings are Zoom)
      - Summary: first 500 chars of AI summary
      - Source: "Fireflies"
      - Fireflies ID: the transcript ID
      - Transcript Link: relative path to transcript file
   f. Update People row: set Last Contacted = meeting date (only if more recent)
3. Log results: X new interactions created, Y skipped (already synced), Z new people created

Schedule: Run after each Fireflies sync. Either:

Add as a second step in run-fireflies-sync.py (call the script after Claude Code sync completes)
Or run as a separate scheduled task (every hour at :15, giving the Fireflies sync time to finish)

Flags:

--plan — dry run, show what would be created
--run — execute
--backfill — process ALL existing transcripts (not just recent ones)

3D. Backfill Existing Transcripts

After the sync script works for new meetings, run once with --backfill:

Processes all ~19 existing summary files (12 meetings + 7 voice notes)
Creates Interactions rows for each
Links to People where possible (name matching)
Gives instant meeting history in the CRM

This is a one-time operation. After this, the hourly sync handles new meetings.

3E. Name Matching Strategy

The trickiest part is linking Fireflies participants to People rows:

Filter out Dmitri — exclude participants matching any of: “Dmitri Sunshine”, “Dmitri Zasage”, or first name “Dmitri” (case-insensitive). Fireflies uses “Dmitri Sunshine” but other contexts use “Dmitri Zasage”.
Build a lookup cache — pull all People rows into a Python dict (179 rows = trivial to cache)
Exact name match — participants[0].name == people_row.Name (case-insensitive, strip whitespace)
Last name match — if exact fails, try matching just the last name (handles “Jonathan” vs “Jonathan Dekle”)
Email match — if Fireflies provides participant email, match against People.Email
No match — create new People row with Status = “New”, Pipeline Source = “Fireflies”
Flag for review — log all auto-created People rows for Dmitri to verify

3F. Validation Checklist

Script connects to Baserow API using credentials from .env
--plan mode shows correct counts without creating anything
--run mode creates Interactions rows with correct Person links
Duplicate Fireflies IDs are skipped (idempotent)
New People rows created for unmatched participants
People.Last Contacted updated correctly
--backfill processes all existing summaries (~19 files across meeting-transcripts/ and voice-notes/)
Script integrated into Fireflies sync schedule (runs after markdown sync)

Phase 4: Formula Fields + Polish

What: Add computed fields that make the CRM self-maintaining + quality-of-life improvements Effort: ~1 hour Dependencies: Phases 1-3 complete, some data in the system

4A. Formula Fields on People Table

Formula Field	Formula Logic	Purpose
Days Since Contact	`if(isblank(field('Last Contacted')), 999, date_diff('day', field('Last Contacted'), today()))`	Instantly see who’s going cold
Contact Status Label	`if(field('Days Since Contact') > 60, '🔴 Cold', if(field('Days Since Contact') > 30, '🟡 Cooling', if(field('Days Since Contact') > 14, '🟢 Warm', '✅ Active')))`	Visual indicator (text-based since no row coloring on free)
Organization Name (lookup)	Use a dedicated Lookup field type (not formula) — pull Name from linked Organization	Show org name directly on People grid without clicking
Full Context	`concat(field('Name'), ' — ', if(isblank(field('Title')), '', concat(field('Title'), ' at ')), join(lookup('Organization', 'Name'), ', '))`	Quick context: “Jonathan Dekle — CEO at Acme Corp”

Formula syntax notes (important for execution):

field('Name') is correct Baserow syntax (not {Name} like Airtable)
isblank() has no underscore (NOT is_blank())
lookup() returns an array even for single links — always wrap in join() when using in concat()
date_diff('day', start, end) uses string literal units: 'day', 'month', 'year', 'hour', 'minute', 'seconds'
Emoji characters (🔴 🟡 🟢 ✅ ⚠️) can be pasted directly into formula strings and work correctly
Equality operator is = not == — Baserow formula syntax uses single equals for comparison (discovered during Phase 4 execution when == caused 400 Bad Request)
created_on is a Baserow auto-field not returned by the fields API — cannot be referenced in view sorts/filters via API; must be configured manually in the UI

4B. Formula Fields on Deals Table

Formula Field	Formula Logic	Purpose
Weighted Value	`field('Estimated Value') * field('Probability') / 100`	Expected revenue
Days in Stage	`date_diff('day', field('Last Activity'), today())`	How long has this deal been stagnant
Stage Alert	`if(field('Days in Stage') > 14, '⚠️ STALE', '')`	Flag stale deals

4C. Formula Fields on Tasks Table

Formula Field	Formula Logic	Purpose
Days Until Due	`date_diff('day', today(), field('Due Date'))`	Negative = overdue
Urgency Label	`if(field('Days Until Due') < 0, '🔴 OVERDUE', if(field('Days Until Due') == 0, '🟡 TODAY', if(field('Days Until Due') <= 3, '🟢 Soon', '')))`	Visual urgency

4D. Webhooks (Lightweight Automation)

Set up these webhooks for future use (or immediate use with simple Python webhook receivers):

Table	Event	What It Enables
Interactions	Row created	Auto-update People.Last Contacted (if not done by script)
Deals	Row updated	Track stage changes for reporting
Tasks	Row updated	Mark tasks done → could trigger next workflow

Webhook setup: Table settings → Webhooks → Add webhook → set URL, events, and headers.

For now, webhooks can point to a simple Python Flask/FastAPI server running locally, or just be configured and left for later when n8n is added.

4E. Validation Checklist

All formula fields compute correctly (test with real data)
Days Since Contact shows accurate values
Weighted Value calculates correctly on Deals
Urgency labels display on Tasks
At least 1 webhook configured and verified (test with webhook.site or similar)

Community Plugins Assessment (Free Only)

The Reality

Baserow’s community plugin ecosystem is small and experimental:

No plugin marketplace — discovery is manual (GitHub search, community forum)
Installation requires building a custom Docker image — not a 1-click process
Plugin API is experimental and subject to breaking changes
Baserow’s own team does not officially support community plugins
No widely-adopted CRM-specific plugins exist

Plugins Evaluated

Plugin	Source	What It Does	Verdict	Why
BaseCharts	GitHub (soon/basecharts)	Chart widgets in views	SKIP	Low maintenance activity. Not worth the Docker image rebuild risk for charts we can generate externally.
BaseScript	GitHub (Raudius/basescript)	In-browser JavaScript execution	SKIP	Proof-of-concept quality. External Python scripts are more powerful and maintainable.
Baserow geo	Community	Map/geography views	SKIP	Premature. Territory mapping isn’t a current need.
Custom formula functions	Plugin API	Add custom formula functions	MAYBE LATER	Could add AI-powered formula functions (call OpenAI for lead scoring, etc.) but premature now.

Recommendation: Zero Plugins for Now

The 80/20 for free Baserow is:

Great relational design (link_row, lookup, rollup, formula)
Smart views (grouped grids, filtered views, gallery, forms)
External Python scripts for automation (Fireflies sync, data import, AI enrichment)
Webhooks for event-driven triggers
REST API for programmatic access

None of these require plugins. The built-in features of free Baserow, used well, deliver a better CRM experience than a plugin-laden setup.

When to reconsider plugins:

When Baserow’s plugin ecosystem matures (target: late 2026)
When a specific workflow cannot be solved by external scripts
When you find a plugin with 100+ GitHub stars and recent commits

Files That Will Be Created/Modified

New Files

File	Purpose
`solanasis-scripts/baserow/schema-upgrade-v2.py`	Create new tables, add fields, migrate Meeting Notes → Interactions
`solanasis-scripts/baserow/sync-fireflies-to-baserow.py`	Sync Fireflies markdown transcripts into Baserow Interactions

Modified Files

File	Change
`solanasis-docs/scheduled-tasks/scripts/run-fireflies-sync.py`	Add call to `sync-fireflies-to-baserow.py` after Claude Code sync
`solanasis-docs/meeting-notes/fireflies-sync-status.md`	Add Baserow sync counts

Reference Files (Read, Don’t Modify)

File	Why Read It
`solanasis-scripts/.env`	Baserow credentials
`solanasis-scripts/foundation-pipeline/migrate_to_baserow.py`	Pattern for Baserow API Python client
`solanasis-scripts/baserow/src/config/tables.ts`	Current schema definition (reference)
`solanasis-scripts/baserow/src/lib/baserow-client.ts`	TypeScript client (reference for API patterns)
`solanasis-docs/scheduled-tasks/01-fireflies-sync-prompt.md`	Fireflies sync configuration

Continuation Prompt for Fresh Session

Copy this into a new Claude Code session to start execution:

We're upgrading Solanasis's self-hosted Baserow CRM. Read these files for full context:

1. `solanasis-docs/reference/deep-plan-baserow-80-20-execution-2026-03-24.md` — THE PLAN (follow this exactly, including the Validation Results section at the top)
2. `solanasis-scripts/.env` — Baserow credentials
3. `solanasis-scripts/foundation-pipeline/migrate_to_baserow.py` — Reference for Baserow API patterns in Python
4. `solanasis-scripts/baserow/src/config/tables.ts` — Current schema

CONTEXT:
- Self-hosted Baserow at baserow.solanasis.com, Database ID: 54
- Free tier only (no Premium)
- Current tables: Tag, Location, Organization, People (179 rows), Meeting Notes (37 rows)
- Plus pipeline tables: Foundation Prospects, fCTO Partners, MSP Prospects
- Plan has been VALIDATED — see Validation Results section for 6 issues that were found and corrected

CRITICAL REQUIREMENTS FROM VALIDATION:
- Script MUST be idempotent (check before create, skip if exists)
- Fireflies name filter must exclude both "Dmitri Sunshine" and "Dmitri Zasage"
- Fireflies backfill scans BOTH meeting-transcripts/ AND voice-notes/ folders (~19 files, not 70)
- Use `isblank()` (no underscore) in formulas, NOT `is_blank()`
- Phase 2 views/forms should be created via API script, not manually

START WITH PHASE 1 from the plan:
1. Create new tables: Interactions, Deals, Tasks
2. Add new fields to People and Organization tables
3. Migrate Meeting Notes → Interactions (37 rows)
4. Validate everything works

Use Python for all scripts. Follow the existing pattern from migrate_to_baserow.py.
Create `solanasis-scripts/baserow/schema-upgrade-v2.py` with --plan and --run flags.

Deferred Items (Documented for Future)

Item	Status	When to Revisit
n8n automation	DEFERRED	When CRM has been in daily use for 2+ weeks and friction points are clear
Apollo.io import	DEFERRED	Separate project — build `import-apollo.py` when Apollo exports are ready
Guild CRM sharing	VISION	Long-term concept. Do not design schema for multi-tenancy now.
Baserow Premium	DEFERRED	Reconsider when 2+ people use the CRM or kanban becomes essential
Community plugins	DEFERRED	Reconsider when plugin ecosystem matures (late 2026)
Pipeline consolidation	DEFERRED	Merge Foundation/fCTO/MSP into unified schema as part of Apollo project
External dashboarding (Metabase)	DEFERRED	Only if formula-based indicators + filtered views aren’t sufficient
Custom ERP/CRM (Frappe)	LONG-TERM	Per `solanasis-custom-erp-crm-build-plan.md` — Baserow is the bridge

Execution Log

Scripts Created (2026-03-25)

Script	Purpose
`solanasis-scripts/baserow/baserow_client.py`	Shared Baserow API client (JWT + DB token auth)
`solanasis-scripts/baserow/schema-upgrade-v2.py`	Phases 1, 2, 4: schema, views, formulas
`solanasis-scripts/baserow/sync-fireflies-to-baserow.py`	Phase 3: Fireflies → Baserow sync

Phase Execution Status

Phase	Status	Started	Completed	Notes
Phase 1: Schema Upgrade	COMPLETE	2026-03-25	2026-03-25	3 tables created, 11 fields added to existing tables, 38 rows migrated. All 1G checklist items pass.
Phase 2: Views & Forms	COMPLETE	2026-03-25	2026-03-25	31 views + 4 forms created. 2 minor: created_on sort skipped (Baserow auto-field not in fields API). All 2G checklist items pass. Form URLs: Quick Contact=e-Fs19p46snkSv-WNE5kzoAIgTKYtkgMx6ubXTMUa9o, Log Interaction=Gs8DnVdhmGXFQk5XaKrx3ekG9GOtf0BJuUPc7uO7KHs, New Deal=-el6oaL46fEnrX-wf5lcwNu-jrXvHeCXEC7Sxqr_SF4, Quick Task=emGyzqEPBT5N1Noxx9bYZyg2468D8BDO3H8wT3j2Pxs
Phase 3: Fireflies Sync	COMPLETE	2026-03-25	2026-03-25	21 interactions created from 23 files (2 skipped=duplicate IDs), 11 new People auto-created, 0 errors. Idempotency verified. 15/21 have Person links (6 are solo voice notes).
Phase 4: Formula Fields	COMPLETE	2026-03-25	2026-03-25	8 formula fields created: People (Days Since Contact, Contact Status Label, Full Context), Deals (Weighted Value, Days in Stage, Stage Alert), Tasks (Days Until Due, Urgency Label). Fixed: Days Since Contact converted from number→formula type, Urgency Label fixed == to = operator. All formulas verified computing correctly with real data.

Senior Review (2026-03-25)

Verdict: APPROVED WITH NOTES

All 4 phases passed code quality, plan compliance, idempotency, and E2E validation.

Known Issues

#	Issue	Severity	Resolution
1	`Last Contacted` not backfilled from Meeting Notes migration — 38 historical meetings are in Interactions but People records don’t reflect those dates, so those contacts show as “Cold”	Medium	One-time fix script needed: read Interactions where Source=Backfill, find most recent date per Person, update `Last Contacted`
2	`created_on` sort unavailable via fields API — “Recently Added” and “All Tasks” views missing their sort	Low	Manual one-time fix in Baserow UI: open each view, add sort by created_on descending
3	Transcript Link paths use Windows backslashes	Low	Change `str(transcript.relative_to(DOCS_DIR))` to `transcript.relative_to(DOCS_DIR).as_posix()` in `sync-fireflies-to-baserow.py` line 251
4	`ensure_field()` calls `list_fields()` per field instead of caching per table	Low	Minor efficiency — only matters if scripts are re-run frequently. Not worth fixing for one-time schema scripts.

Manual Follow-Up Items

Bookmark 4 form URLs in browser toolbar (per plan section 2F)
Add created_on sort to “Recently Added” and “All Tasks” views in UI
Run backfill script for Last Contacted on historical Meeting Notes contacts (when written)

Table IDs (Live Reference)

Table	ID	Row Count (at execution)
People	267	2,666
Organization	266	66
Interactions	769	59 (38 backfill + 21 Fireflies)
Deals	770	0
Tasks	771	0
Tag	264	51
Location	265	—
Meeting Notes (ARCHIVED)	272	37

Form URLs (Live Reference)

Form	Full URL
Quick Contact	`https://baserow.solanasis.com/form/e-Fs19p46snkSv-WNE5kzoAIgTKYtkgMx6ubXTMUa9o`
Log Interaction	`https://baserow.solanasis.com/form/Gs8DnVdhmGXFQk5XaKrx3ekG9GOtf0BJuUPc7uO7KHs`
New Deal	`https://baserow.solanasis.com/form/-el6oaL46fEnrX-wf5lcwNu-jrXvHeCXEC7Sxqr_SF4`
Quick Task	`https://baserow.solanasis.com/form/emGyzqEPBT5N1Noxx9bYZyg2468D8BDO3H8wT3j2Pxs`

Formula Field Reference

Table	Field	Formula	Computes
People	Days Since Contact	`if(isblank(field('Last Contacted')), 999, date_diff('day', field('Last Contacted'), today()))`	Days since last contact; 999 if never
People	Contact Status Label	`if(field('Days Since Contact') > 60, '🔴 Cold', if(field('Days Since Contact') > 30, '🟡 Cooling', if(field('Days Since Contact') > 14, '🟢 Warm', '✅ Active')))`	Emoji status tier
People	Full Context	`concat(field('Name'), ' — ', if(isblank(field('Title')), '', concat(field('Title'), ' at ')), join(lookup('Organization', 'Name'), ', '))`	Name + title + org summary
Deals	Weighted Value	`field('Estimated Value') * field('Probability') / 100`	Expected revenue
Deals	Days in Stage	`date_diff('day', field('Last Activity'), today())`	Staleness indicator
Deals	Stage Alert	`if(field('Days in Stage') > 14, '⚠️ STALE', '')`	Warning for stale deals
Tasks	Days Until Due	`date_diff('day', today(), field('Due Date'))`	Countdown to deadline
Tasks	Urgency Label	`if(field('Days Until Due') < 0, '🔴 OVERDUE', if(field('Days Until Due') = 0, '🟡 TODAY', if(field('Days Until Due') <= 3, '🟢 Soon', '')))`	Emoji urgency tier

Solanasis Docs

Explorer

deep-plan-baserow-80-20-execution-2026-03-24