nathan/AI-Prototype
1
0
Fork 0
Code Pull Requests Packages Projects Activity

testing out OCR viewer of OCR'd text

Browse Source
This commit is contained in:
nathan
2026-05-09 09:36:30 -04:00
parent 727f4fcc57
commit ba297d6430
19 changed files with 2674 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
prompts/unit-selection.md
View File

@@ -4,7 +4,8 @@
Calgary highlanders, part of 5th Canadian Infantry Brigade, 2nd Canadian Infantry Division.
## Why this unit
Good OCR coverage, high quality PDFs with lots of narrative. Content spans Sep 44 to Nov 44, though the OCR text extends to May 45, nothign is needed past Dec 44.
Good OCR coverage, high quality PDFs with lots of narrative. Content spans Sep 44 to Nov 44,
though the OCR text extends to May 45, nothing is needed past Dec 44.
## Hypothetical customer scenario
Pte. Bloggins, reinforcement who joined the battalion mid September 1944
@@ -15,4 +16,6 @@ Family knows: the regiment, that he was wounded somewhere in "Holland or Germany
in late 1944, the rough timeframe.
## What I expect the report to cover
Fighting in France and Belgium, the advance towards Antwerp, and the Scheldt Battles. Should feel like a personal narrative, outlining the experience of a private during the battles. Try and find any personal narratives that might give the family a sense of what the battles were like, or life was like.
Fighting in France and Belgium, the advance towards Antwerp, and the Scheldt Battles.
Should feel like a personal narrative, outlining the experience of a private during the battles.
Try and find any personal narratives that might give the family a sense of what the battles were like, or life was like.

97
prompts/v1-tier3-report.md Normal file
View File

@@ -0,0 +1,97 @@
# Prompt: Tier 3 Descendant Report — Calgary Highlanders
## Your role
You are helping produce a report for a descendant of a Canadian WWII soldier.
The report draws on the soldier's regimental war diary to reconstruct what
his unit experienced during his service window. You are NOT writing a
generic regimental history — you are writing for one family, grounded in
what they know and shaped around the soldier's likely experience.
## Input
You will be given the war diary OCR for the Calgary Highlanders covering
September 1944 through May 1945. The diary is in .docx form with a three-
column structure:
- Column 1: Date
- Column 2: Narrative entry (the diarist's account of the day)
- Column 3: Comments / page reference back to the original diary
Each row is a day's entry. Some entries are operational (movements, attacks,
casualties), some are administrative (parades, pay, training), some are
human texture (weather, food, civilian encounters, rest). All of these
matter for the report — the human texture is what makes a descendant report
emotionally resonant, not just the operational events.
## The customer's situation (Tier 3 — partial information)
The family knows:
- Their grandfather, Pte. Bill Bloggins, served with the Calgary Highlanders.
- He was a reinforcement who joined the battalion in mid September 1944
during the fighting in France.
- He was wounded in action in early November 1944, somewhere in
"Holland or Germany," during what the family calls "the Scheldt."
- He was evacuated to England and did not return to the unit.
The family does NOT know:
- The exact date he joined or was wounded.
- Specific actions he was personally involved in.
- His company, platoon, or section.
## What to produce
### 1. Narrative report (1,200 – 1,800 words)
A continuous prose narrative covering the unit's experience from mid September 1944 through early November 1944, framed for Bloggins's family.
Structure suggestion (not mandatory):
- Brief opening situating the unit and the moment Bill likely joined.
- The Scheldt fighting as his first weeks with the battalion.
- The static winter on the Maas.
- The Rhineland operations and the period during which Bill was wounded.
- A short closing reflecting on what his service likely looked like.
The voice should be:
- Grounded and specific — name places, dates, named officers, weather,
details from the diary. Avoid generic phrases like "the brave Canadians"
or "in the face of fierce resistance."
- Honest about uncertainty — when the family doesn't know exact dates,
the narrative says so plainly. ("We don't know the exact day Bill
joined, but the battalion was at X during the first week of October...")
- Human as well as operational — include the texture (rations, billets,
civilian encounters, the men's mood) alongside the fighting.
### 2. Proposed events list (10 – 20 events)
A list of events from the service window that should be pinned to the map
for human review. For each event, give:
- Date (from the diary)
- Location as described in the diary (verbatim, including grid references
if present)
- One-sentence description of what happened
- Why this event matters for Bill's story (operational significance,
emotional weight, or representative of daily experience)
- Source page reference from the comments column
Include a mix: major engagements, smaller actions, movements,
representative quiet days. Not every event needs to be a battle.
### 3. Documents and passages of interest (5 – 10 items)
Specific entries or passages from the diary that the family might want to
read in full themselves — things that bring Bill's experience to life,
or that mark turning points, or that contain unusual detail. For each:
- Date and page reference
- Brief description of why it's worth reading
- Only focus on Sep 44 to Nov 44 in the documents. Skip any months after that.
## What NOT to do
- Do NOT invent specifics. If the diary doesn't say where the battalion
was on a given day, say so. If a name isn't in the diary, don't add one.
- Do NOT smooth over gaps in the record with plausible-sounding filler.
- Do NOT write in a generic war-documentary voice. Stay close to the
diary's actual content and tone.
- Do NOT speculate about Bill's personal experiences beyond what the
unit-level diary supports. We don't know if he was scared, brave,
homesick, etc. We know what his battalion did.
- Do NOT pad the narrative to hit the word count. If the diary is thin
for a period, the narrative is thin for that period.
## Output format
Produce the three sections in order, clearly labeled. Use Markdown.
The narrative is prose; the events and documents lists can be structured
as Markdown lists or tables, your choice.****

97
prompts/v2-tier3-report.md Normal file
View File

@@ -0,0 +1,97 @@
# Prompt: Tier 3 Descendant Report — Calgary Highlanders
## Your role
You are helping produce a report for a descendant of a Canadian WWII soldier.
The report draws on the soldier's regimental war diary to reconstruct what
his unit experienced during his service window. You are NOT writing a
generic regimental history — you are writing for one family, grounded in
what they know and shaped around the soldier's likely experience.
## Input
You will be given the war diary OCR for the Calgary Highlanders covering
September 1944 through May 1945. The diary is in .docx form with a three-
column structure:
- Column 1: Date
- Column 2: Narrative entry (the diarist's account of the day)
- Column 3: Comments / page reference back to the original diary
Each row is a day's entry. Some entries are operational (movements, attacks,
casualties), some are administrative (parades, pay, training), some are
human texture (weather, food, civilian encounters, rest). All of these
matter for the report — the human texture is what makes a descendant report
emotionally resonant, not just the operational events.
## The customer's situation (Tier 3 — partial information)
The family knows:
- Their grandfather, Pte. Bill Bloggins, served with the Calgary Highlanders.
- He was a reinforcement who joined the battalion in mid September 1944
during the fighting in France.
- He was wounded in action in early November 1944, somewhere in
"Holland or Germany," during what the family calls "the Scheldt."
- He was evacuated to England and did not return to the unit.
The family does NOT know:
- The exact date he joined or was wounded.
- Specific actions he was personally involved in.
- His company, platoon, or section.
## What to produce
### 1. Narrative report (1,200 – 1,800 words)
A continuous prose narrative covering the unit's experience from mid September 1944 through early November 1944, framed for Bloggins's family.
Structure suggestion (not mandatory):
- Brief opening situating the unit and the moment Bill likely joined.
- The Scheldt fighting as his first weeks with the battalion.
- The static winter on the Maas.
- The Rhineland operations and the period during which Bill was wounded.
- A short closing reflecting on what his service likely looked like.
The voice should be:
- Grounded and specific — name places, dates, named officers, weather,
details from the diary. Avoid generic phrases like "the brave Canadians"
or "in the face of fierce resistance."
- Honest about uncertainty — when the family doesn't know exact dates,
the narrative says so plainly. ("We don't know the exact day Bill
joined, but the battalion was at X during the first week of October...")
- Human as well as operational — include the texture (rations, billets,
civilian encounters, the men's mood) alongside the fighting.
### 2. Proposed events list (10 – 20 events)
A list of events from the service window that should be pinned to the map
for human review. For each event, give:
- Date (from the diary)
- Location as described in the diary (verbatim, including grid references
if present)
- One-sentence description of what happened
- Why this event matters for Bill's story (operational significance,
emotional weight, or representative of daily experience)
- Source page reference from the comments column
Include a mix: major engagements, smaller actions, movements,
representative quiet days. Not every event needs to be a battle.
### 3. Documents and passages of interest (5 – 10 items)
Specific entries or passages from the diary that the family might want to
read in full themselves — things that bring Bill's experience to life,
or that mark turning points, or that contain unusual detail. For each:
- Date and page reference
- Brief description of why it's worth reading
- Only focus on Sep 44 to Nov 44 in the documents. Skip any months after that.
## What NOT to do
- Do NOT invent specifics. If the diary doesn't say where the battalion
was on a given day, say so. If a name isn't in the diary, don't add one.
- Do NOT smooth over gaps in the record with plausible-sounding filler.
- Do NOT write in a generic war-documentary voice. Stay close to the
diary's actual content and tone.
- Do NOT speculate about Bill's personal experiences beyond what the
unit-level diary supports. We don't know if he was scared, brave,
homesick, etc. We know what his battalion did.
- Do NOT pad the narrative to hit the word count. If the diary is thin
for a period, the narrative is thin for that period.
## Output format
Produce the three sections in order, clearly labeled. Use Markdown.
The narrative is prose; the events and documents lists can be structured
as Markdown lists or tables, your choice.****