Dataset visualizations

Author: mangelhc

.DS_Store

@@ -0,0 +1 @@
+*.mp4 filter=lfs diff=lfs merge=lfs -text

.gitattributes

@@ -1,7 +1,4 @@
-__pycache__/
-*.pyc
+
 .venv/
-.idea/
-.vscode/
-*.bsp
-*.tle
+**/__pycache__/
+.DS_Store

.gitignore

@@ -123,11 +123,17 @@ task_generation:
            * VNIR -> Explain the goal as "multispectral vegetation analysis" to assess "chlorophyll absorption level" or calculate a vegetation "NDVI index" and surface reflectance.
            * NIR -> Explain the goal as "near-infrared reflection" for "water body boundary mapping" or "soil moisture assessment" to discriminate gradients.
 
-      4. RELATIVE TIMEFRAMES (Interpret keys 'remaining_time' and 'deadline_local'):
-         - Do NOT under any circumstances output raw UTC timestamps, Unix epochs, digits, raw numbers, or countdown metrics.
-         - Do NOT use composite time expressions that sum hours and days (e.g., do NOT write phrases like "in 3 days and 5 hours", "with 12 hours remaining", or "48 hours from now"). This is strictly forbidden.
-         - You must deliver the temporal constraint using single, fluid, human-centric relative phrasing. Fuse the macro day and the local diurnal hour into a single colloquial block. Speak naturally like an operator on a radio link: integrate time expressions by rounding the constraint into a clean target window such as "tomorrow morning", "this afternoon", "the day after tomorrow, specifically overnight", "in three days during the evening", or "sometime today around mid-day". Let the language flow without structural or robotic rigidity.
-
+      4. RELATIVE TIMEFRAMES (Interpret keys 'target_day' and 'target_diurnal_period'):
+              - Do NOT output raw UTC timestamps, digits, or countdown metrics.
+              - CRITICAL CONTEXT: The temporal parameters represent a strict operational DEADLINE to deliver the data products to the user, NOT the orbital capture window. Your text must reflect this urgency naturally.
+              - You are strictly forbidden from altering or extending this deadline. You must weave 'target_day' and 'target_diurnal_period' into a fluid conversational phrase that explicitly marks the upper limit for data availability using phrases like:
+                * "...needed by [target_day] [target_diurnal_period] at the latest."
+                * "...with a strict delivery deadline set for [target_day] [target_diurnal_period]."
+                * "...ensuring the products are ingested into the pipeline before [target_day] [target_diurnal_period]."
+              
+              Operational vocabulary alignment:
+              - If 'target_diurnal_period' is "overnight" -> The data must be ready by tomorrow's overnight/nighttime block. Do NOT write "afternoon" or "morning" as it would violate the deadline constraint.  
+      
       Target task context JSON:
       {tasks_dataset}
 
@@ -138,4 +144,3 @@ task_generation:
 
 paths:
   gs_file_path: "data/ground_station.csv"
-  sat_file_path: "data/satellite_tle.csv"

config.yaml

@@ -123,11 +123,17 @@ task_generation:
            * VNIR -> Explain the goal as "multispectral vegetation analysis" to assess "chlorophyll absorption level" or calculate a vegetation "NDVI index" and surface reflectance.
            * NIR -> Explain the goal as "near-infrared reflection" for "water body boundary mapping" or "soil moisture assessment" to discriminate gradients.
 
-      4. RELATIVE TIMEFRAMES (Interpret keys 'remaining_time' and 'deadline_local'):
-         - Do NOT under any circumstances output raw UTC timestamps, Unix epochs, digits, raw numbers, or countdown metrics.
-         - Do NOT use composite time expressions that sum hours and days (e.g., do NOT write phrases like "in 3 days and 5 hours", "with 12 hours remaining", or "48 hours from now"). This is strictly forbidden.
-         - You must deliver the temporal constraint using single, fluid, human-centric relative phrasing. Fuse the macro day and the local diurnal hour into a single colloquial block. Speak naturally like an operator on a radio link: integrate time expressions by rounding the constraint into a clean target window such as "tomorrow morning", "this afternoon", "the day after tomorrow, specifically overnight", "in three days during the evening", or "sometime today around mid-day". Let the language flow without structural or robotic rigidity.
-
+      4. RELATIVE TIMEFRAMES (Interpret keys 'target_day' and 'target_diurnal_period'):
+              - Do NOT output raw UTC timestamps, digits, or countdown metrics.
+              - CRITICAL CONTEXT: The temporal parameters represent a strict operational DEADLINE to deliver the data products to the user, NOT the orbital capture window. Your text must reflect this urgency naturally.
+              - You are strictly forbidden from altering or extending this deadline. You must weave 'target_day' and 'target_diurnal_period' into a fluid conversational phrase that explicitly marks the upper limit for data availability using phrases like:
+                * "...needed by [target_day] [target_diurnal_period] at the latest."
+                * "...with a strict delivery deadline set for [target_day] [target_diurnal_period]."
+                * "...ensuring the products are ingested into the pipeline before [target_day] [target_diurnal_period]."
+              
+              Operational vocabulary alignment:
+              - If 'target_diurnal_period' is "overnight" -> The data must be ready by tomorrow's overnight/nighttime block. Do NOT write "afternoon" or "morning" as it would violate the deadline constraint.  
+      
       Target task context JSON:
       {tasks_dataset}
 
@@ -138,4 +144,3 @@ task_generation:
 
 paths:
   gs_file_path: "data/ground_station.csv"
-  sat_file_path: "data/satellite_tle.csv"

data/constellation_dataset_gemma2_27b_v2/config.yaml

@@ -1 +1 @@
-This is a low priority importance request for a non-essential background filler task. We need an opportunistic capture of multispectral vegetation analysis data over El Paujil, Colombia sometime tomorrow during the afternoon.  Feel free to drop or reschedule at convenience.
+This is a low priority importance request for a VNIR multispectral vegetation analysis scan over El Paujil, Colombia.  We're looking to capture some opportunistic data for chlorophyll absorption level assessment and NDVI index calculation.  The products can be ingested into the pipeline at your earliest convenience, but ideally before tomorrow evening.

data/constellation_dataset_gemma2_27b_v2/scenario_1/ollama_prompt_TASK_GEN_001.txt

@@ -1 +1 @@
-This is an absolute institutional priority tasking. We require a critical payload capture of thermal mapping data over Calvi, France. Schedule this mandatory directive for tomorrow morning.
+This is an absolute institutional priority tasking. We require a critical payload capture of the Calvi region in France using thermal mapping.  The data products are needed by tomorrow morning at the latest, ensuring they are ingested into the pipeline before then.

data/constellation_dataset_gemma2_27b_v2/scenario_1/ollama_prompt_TASK_GEN_002.txt

@@ -1 +1 @@
-This is a low priority importance task for opportunistic capture. I need a thermal mapping of Ecuatorial, Indonesia sometime the day after tomorrow during the overnight hours.
+This is a low priority importance request for a thermal mapping survey over Ecuatorial, Indonesia. We need an overnight heat signature capture of the region by the day after tomorrow at the latest. This is a non-essential background filler task and can be dropped or rescheduled at convenience.

data/constellation_dataset_gemma2_27b_v2/scenario_1/ollama_prompt_TASK_GEN_003.txt

@@ -1 +1 @@
-This is a low priority importance request for opportunistic capture. We need a multispectral vegetation analysis snapshot of Ibagué, Colombia sometime the day after tomorrow around mid-day.  Feel free to drop or reschedule this at your convenience; it's a non-essential background filler task.
+This is a low priority importance request for a multispectral vegetation analysis survey over Ibagué, Colombia. We're looking for an opportunistic capture of the area using near-infrared reflection to assess chlorophyll absorption levels and calculate a vegetation NDVI index.  Please drop or reschedule at your convenience, treating this as a background survey task. The products need to be ingested into the pipeline before the day after tomorrow during the afternoon at the latest.

data/constellation_dataset_gemma2_27b_v2/scenario_1/ollama_prompt_TASK_GEN_004.txt

@@ -1 +1 @@
-This is a critical payload capture and an absolute institutional priority. We require a high-resolution optical snapshot of Norte, sometime tomorrow during the afternoon.  Please ensure this task is un-preemptable and executed with the highest urgency.
+This is an absolute institutional priority tasking. We require a high-resolution optical snapshot of Ponte-Novo, France.  The capture must be completed and the products ingested into the pipeline before tomorrow during the afternoon at the latest. This is a critical payload capture and represents a mandatory directive.