as we speak deal with how a number of brokers coordinate whereas choosing instruments from a predefined toolbox. Whereas efficient, this design quietly assumes that the instruments required for a activity are recognized prematurely.
Let’s problem that assumption on this article.
Trendy LLMs can already decompose complicated objectives into multi-step plans, purpose about execution order, and generate appropriate, executable code. In different phrases, they don’t simply use instruments — they’ll create them.
This text explores utilizing a plan–code–execute agentic structure to construct an end-to-end Explainability Agent, the place instruments are generated on demand, illustrated by way of an actual world state of affairs.
The structure proposed right here is well adaptable to different eventualities as effectively.
The Situation
We will likely be explaining the outcomes of a Graph Neural Community (GNN) mannequin for demand forecasting, that I’ve explored in my earlier article. We have now the next information indicators to clarify:
- The Spatial options of the nodes (SKUs).
- The Temporal options (Gross sales Order, Manufacturing and so forth) added to the graph nodes.
- Outcomes from a graph neural community mannequin educated to forecast demand for a SKU.
- Outcomes from a GNN educated to detect anomalies in manufacturing.
All of those generated a sequence of analyses, outcomes and charts which must be defined holistically, from a planner’s perspective, that’s grounded in information and area understanding.
Explainability Agent Structure
We’ll create a plan-code-execute agentic workflow. To have granular management, interpretability and adaptability in software technology, we’ll construct totally customized brokers relatively than working by way of the abstractions of one of many well-known orchestration frameworks. Following brokers are created:
1. The Analyst Agent
We’d like an analyst agent which might be the eyes of the appliance. Its job is to Uncover the surroundings (Uncooked Knowledge, Mannequin Outcomes, Earlier Outputs) which the remainder of the appliance will use. It prevents hallucinations such because the planner agent assuming the gross sales file is called “gross sales.csv” when the right file identify is “Gross sales Order.csv” within the dataset.
Along with simply scanning what’s there, it additionally opens the information and creates a dynamic schema. As an example, Gross sales Order.csv has columns [Date, Amount] and demand_vs_production.csv has [Predicted, Actual]. This varieties the context for the planning agent.
2. The Planner
The Planner receives the person’s high-level purpose and the Analyst’s schema. It capabilities because the Mind of the appliance. Its job is to decompose the request right into a dependency graph of steps. To try this, it first infers intent from the person request. If the person asks for “Accuracy,” it is aware of to focus on the Forecast Outcomes folder. In the event that they ask for “Gross sales historical past,” it targets Uncooked Knowledge. The output is a JSON plan which has steps clearly describing the duty and likewise as well as a flag known as CODE or TEXT. Instruments are generated by the Coder agent just for the CODE steps.
Pattern output: A JSON plan.
- Step 1: Load Graph information and discover neighbors of SOS001L12P. (CODE)
- Step 2: Combination temporal gross sales vs manufacturing. (CODE)
- Step 3: Calculate WAPE. (CODE)
- Step 4: Put together abstract (TEXT)
3. The Coder
The Coder is the arms of the appliance, which does the majority of the work. The Coder receives every step from the plan and the schema. For each plan CODE step, it writes a standalone, executable Python script.
It handles library imports (pandas, networkx), absolute paths, and error catching. It creates self sustaining software program, not simply code snippets.
4. The Executor
The Executor runs the generated script in a managed surroundings. So as to do this, it captures stdout and stderr. If a script fails (e.g., FileNotFound), the logs present the suggestions loop for potential self-correction (although as a result of grounding supplied by the Analyst’s schema this doesn’t occur).
5. The Reporter
Lastly, the Reporter reads the logs and artifacts (CSVs, PNGs) produced by the scripts and solutions the person’s authentic query, within the tone, depth and persona required by the person. It capabilities because the Voice of the appliance.
The brokers are known as in sequence utilizing a easy orchestration script (fundamental.py). A config.py incorporates the trail of the dataset folder containing the nodes and edges data, and to the outcomes folders of the demand forecasting and manufacturing anomaly detection fashions. Every of the brokers and the orchestration script is lower than 100 traces of code. The LLM used is gemini-2.5-flash.
The primary errors encountered throughout testing was the coder agent referring an information supply that’s not current (hallucination), for which I strengthened the coder immediate to strictly comply with the plan step context. Additionally, the analyst agent was initially not there, and was added as a grounding step earlier than the planner to forestall hallucination.
Right here is the planner agent code for reference
import json
from ..llm.shopper import GeminiClient
class Planner:
def __init__(self):
self.llm = GeminiClient()
def create_plan(self, user_prompt, dataset_context=""):
"""Decomposes person immediate into execution steps."""
system_prompt = f"""
You're a Senior Challenge Supervisor for a Knowledge Science group.
Your purpose is to interrupt down a high-level person request into an inventory of particular, executable steps.
--- DATASET AVAILABLE (Use EXACT Filenames) ---
{dataset_context}
-----------------------------------------------
--- DATASET GUIDELINES ---
- The dataset incorporates each 'Uncooked Knowledge' (Historic) and 'Mannequin Outcomes' (Predictions/Anomalies).
- CAREFULLY infer which supply is required primarily based on the person's purpose.
- Instance: "Gross sales Evaluation" -> Uncooked Knowledge. "Forecast Accuracy" -> Mannequin Outcomes.
Output format: JSON listing of objects with fields:
- step_id: int
- identify: str (quick title)
- description: str (detailed instruction for the coder)
- sort: "CODE" (if it requires writing a script) or "TEXT" (if only a abstract)
--- SCOPE RESTRICTION ---
- DO NOT USE Machine Studying (LinearRegression, Coaching, Random Forest).
- USE ONLY Statistical Evaluation (Imply, Median, Std Dev, Z-Rating, Rolling Averages).
- FOCUS on Knowledge Visualization (Line plots, Histograms).
- If the person asks for "forecast" or "accuracy", PREFER WAPE (Weighted MAPE) over customary MAPE.
Instance:
Person: "Analyze SKU001"
Output:
[
{{"step_id": 1, "name": "Load Data", "description": "Load Nodes.csv and filter for SKU001 to get basic info.", "type": "CODE"}},
{{"step_id": 2, "name": "Temporal Analysis", "description": "Load Production.csv and Sales.csv. Calculate monthly volume and trend for SKU001.", "type": "CODE"}}
]
"""
full_prompt = f"{system_prompt}nnUser Request: {user_prompt}nJSON Plan:"
response = self.llm.generate_content(full_prompt)
# primary cleanup for JSON markdown blocks
clean_response = response.exchange("```json", "").exchange("```", "").strip()
print(f"n[DEBUG] Uncooked Planner Response:n{response}n") # Added for debugging
attempt:
return json.masses(clean_response)
besides json.JSONDecodeError:
print(f"Didn't parse plan: {response}")
return []Experiment Outcomes
We’ll deal with the end-to-end explainability for SKU SOS001L12P. That is the node with the utmost gross sales quantity. Planners and administration can be all in favour of understanding points early with the intention to fulfill demand and preserve service high quality.
We’ll construct the explainability in 4 steps:
- Analyse the shared relations of this SKU with its neighbors within the provide community.
- Analyse the development in gross sales quantity.
- Correlate with the manufacturing sign, notice anomalies.
- Create a complete report combining with demand forecast high quality to clarify the well being of this SKU.
Spatial Evaluation
The purpose set for the agent was
“carry out an intensive evaluation of the spatial construction round node SOS001L12P and provides me a visualization of the graph
edges with labels connecting this SKU with its neighbors”
The planner agent created a 5-step plan for loading node and edge information, extracting neighbors and edges attributes, making ready a graph visualization with labels and making a report for the spatial construction.
Right here is the plan
[
{
"step_id": 1,
"name": "Load Node and Edge Data",
"description": "Load 'Nodes.csv' to confirm 'SOS001L12P' as a valid node. Load all edge datasets: 'Edges (Plant).csv', 'Edges (Product Group).csv', 'Edges (Product Sub-Group).csv', and 'Edges (Storage Location).csv'. These files contain the structural information needed for spatial analysis.",
"type": "CODE"
},
{
"step_id": 2,
"name": "Extract Direct Neighbors and Edge Attributes",
"description": "Iterate through each loaded edge dataset. For each dataset, filter rows where 'SOS001L12P' is present in either the 'node1' or 'node2' column. For each matching record, identify the neighboring node and extract the specific edge attribute that defines the relationship (e.g., 'Plant' from 'Edges (Plant).csv', 'GroupCode' from 'Edges (Product Group).csv', 'SubGroupCode' from 'Edges (Product Sub-Group).csv', 'Storage Location' from 'Edges (Storage Location).csv'). Consolidate all unique neighbors and their corresponding labeled relationships into a single structure.",
"type": "CODE"
},
{
"step_id": 3,
"name": "Prepare Graph for Visualization",
"description": "Construct a graph object (e.g., using NetworkX) with 'SOS001L12P' as the central node. Add all identified direct neighbors as nodes and create edges between 'SOS001L12P' and each neighbor. Assign the extracted edge attributes (e.g., 'Plant: Plant A', 'GroupCode: GRP1') as labels to these edges for clarity in the visualization.",
"type": "CODE"
},
{
"step_id": 4,
"name": "Visualize Graph Edges with Labels",
"description": "Generate a clear graph visualization of the spatial structure around 'SOS001L12P'. The plot should display 'SOS001L12P' prominently in the center, with its direct neighbors connected by edges. Each edge must be explicitly labeled with the type of relationship and its specific attribute value (e.g., 'Plant: P001', 'Product Group: PG002'). Ensure appropriate layout and styling for readability.",
"type": "CODE"
},
{
"step_id": 5,
"name": "Summarize Spatial Connectivity",
"description": "Provide a textual summary of the spatial structure around 'SOS001L12P'. Describe the types of entities it is connected to (Plants, Product Groups, Product Sub-Groups, Storage Locations), list the specific neighbors identified, and briefly explain the nature of these connections based on the edge labels.",
"type": "TEXT"
}
]The coder agent created 4 instruments, one for every of the CODE steps
And the first output is the graph of this SKU’s relations with its neighbors with all of the relations labeled, reminiscent of shared vegetation, storage places, product teams and so forth.
And a short report is generated of the findings which notes:
“Neighbor Relationships: The evaluation recognized 39 distinct relationships connecting SOS001L12P to 39 distinctive neighboring nodes.”
Full report right here
An intensive evaluation of the spatial construction round node `SOS001L12P` has been accomplished.
**Evaluation Findings:**
1. **Node Validation:** The node `SOS001L12P` was confirmed as a legitimate node inside the grasp information.
2. **Neighbor Relationships:** The evaluation recognized 39 distinct relationships connecting `SOS001L12P` to 39 distinctive neighboring nodes. These relationships are categorized by `RelationshipType` and `RelationshipValue`:
* **Plant:** `SOS001L12P` is linked to a number of neighbors through numerous 'Plant' codes (e.g., 2103, 2111, 2112, 2114, 2115, 2116, 2117, 2118, 2119, 2120, 2121, 2122).
* **GroupCode:** `SOS001L12P` shares 'GroupCode' 'S' with neighbors like `SOS008L02P`, `SOS005L04P`, `SOS003L04P`, `SOS002L09P`, `SOS500M24P`, `SOS250M48P`, `SOP001L12P`, `SO0005L04P`, `SO0002L09P`, `SO0001L12P`, `SO0500M24P`, `SE500G24P`, and `SE200G24P`.
* **SubGroupCode:** `SOS001L12P` is linked by 'SubGroupCode' 'SOS' to nodes reminiscent of `SOS008L02P`, `SOS005L04P`, `SOS003L04P`, `SOS002L09P`, `SOS500M24P`, and `SOS250M48P`.
* **Storage Location:** `SOS001L12P` is related to its neighbors by way of numerous 'Storage Location' codes (e.g., 330.0, 1130.0, 1230.0, 1430.0, 1530.0, 1630.0, 1730.0, 1830.0, 1930.0, 2030.0, 2130.0, 2230.0).
The complete listing of neighbors and their particular relationships might be present in `C:/Auto-agent/workspace/outputSOS001L12P_neighbors_relationships.csv`.
**Visualization:**
A graph visualization has been generated, centering on `SOS001L12P` and displaying all its direct neighbors and their connecting edges with labels indicating the `RelationshipType` and `RelationshipValue`. The ensuing graph incorporates 40 nodes (together with `SOS001L12P`) and 39 edges.
The visualization is out there at: `C:/Auto-agent/workspace/outputSOS001L12P_neighbor_graph.png`Evaluation of Gross sales Order (Demand) sign
Since we have an interest within the accuracy of the demand forecasting mannequin, it is very important analyse the gross sales traits. The purpose was set to:
“carry out evaluation of gross sales order temporal sign of SOS001L12P noting the development, outliers and any abnormalities in quantity traits”
The planner generated a 6-step plan to load and course of gross sales information, create 7-day and 30-day rolling averages, calculate outliers, visualize traits and put together a report.
The plan is right here
[
{
"step_id": 1,
"name": "Load Sales Order Data for SOS001L12P",
"description": "Load the 'Sales Order.csv' file from C:AnomalyGNNdatasetRaw DatasetTemporal DataUnitSales Order.csv. Focus on the 'Date' and 'SOS001L12P' columns, treating 'SOS001L12P' as the sales volume.",
"type": "CODE"
},
{
"step_id": 2,
"name": "Preprocess Temporal Data",
"description": "Convert the 'Date' column to datetime objects and set it as the DataFrame index. Ensure the sales volume column (SOS001L12P) is numerical. Handle any missing values, potentially by filling with 0 or forward/backward fill if appropriate for sales data (specify chosen method).",
"type": "CODE"
},
{
"step_id": 3,
"name": "Analyze Sales Volume Trends",
"description": "Calculate 7-day and 30-day rolling averages for the 'SOS001L12P' sales volume. These will help identify short-term and long-term trends in the sales signal. Store these as new columns in the DataFrame.",
"type": "CODE"
},
{
"step_id": 4,
"name": "Detect Outliers using Z-Score",
"description": "Calculate the Z-score for the 'SOS001L12P' sales volume. Identify potential outliers by marking data points where the absolute Z-score exceeds a predefined threshold (e.g., 2 or 3 standard deviations). Store the Z-score and an 'Is_Outlier' boolean column.",
"type": "CODE"
},
{
"step_id": 5,
"name": "Visualize Sales Order Temporal Signal",
"description": "Create a time-series line plot showing the 'SOS001L12P' sales volume over time. Overlay the 7-day and 30-day rolling averages. Highlight the identified outliers on the plot to visually represent abnormalities in volume trends.",
"type": "CODE"
},
{
"step_id": 6,
"name": "Summarize Trends and Abnormalities",
"description": "Provide a textual summary of the observed trends (e.g., seasonality, growth/decline), the frequency and magnitude of identified outliers, and any other notable abnormalities or shifts in the sales volume patterns for SOS001L12P. This summary should interpret the visualizations and statistical measures.",
"type": "TEXT"
}
]The coder agent created 5 instruments and the first output of the pipeline is the next development. The notable commentary is a downward development within the second half of the length which may very well be a trigger for concern that must be investigated.
The evaluation report additionally notes this declining development.
“Persistent Decline: The constant lower in each day by day gross sales and the rolling averages (7-day and 30-day) from January to August clearly signifies a persistent decline in gross sales quantity for SOS001L12P.“
Right here is the total report
**Evaluation of SOS001L12P Gross sales Order Temporal Sign**
This evaluation covers gross sales information for product SOS001L12P from January 1, 2023, to August 9, 2023.
**1. Quantity Development Through the Length:**
The gross sales quantity for SOS001L12P reveals a transparent and protracted declining development all through the analyzed interval.
* In early January 2023, day by day gross sales volumes had been excessive, with particular person days reaching as much as 21,394 items. The 30-day rolling common was roughly 12,629 by January tenth.
* By early August 2023, the gross sales quantity had considerably decreased. The 30-day rolling common plummeted to roughly 5,165 by August ninth, indicating a considerable drop in constant gross sales exercise over the eight-month interval.
**2. Persistent Declines and Different Points:**
* **Persistent Decline:** The constant lower in each day by day gross sales and the rolling averages (7-day and 30-day) from January to August clearly signifies a persistent decline in gross sales quantity for SOS001L12P.
* **Anomalies:** The evaluation recognized 14 particular days categorized as anomalies. These had been characterised by unusually low gross sales volumes, both because of a Z-score beneath -2.0 or three consecutive days of detrimental proportion change. Notable extreme drops embrace:
* January 4, 2023: Gross sales dropped to 10,920.08.
* March 8, 2023: Gross sales recorded an exceptionally low worth of 200.00.
* **Zero Gross sales:** In direction of the top of the analyzed interval, there have been situations of zero gross sales recorded (e.g., August eighth and ninth), additional exacerbating the declining development and highlighting vital points in gross sales efficiency.Evaluation of Manufacturing Sign
You will need to correlate the manufacturing anomalies report back to see if there’s a causal indication as to why the gross sales could be declining. To do that, we set the next purpose for the agent:
“carry out evaluation of the manufacturing temporal sign of SOS001L12P ,declaring the development, outliers and any abnormalities in quantity traits“
A 7-step plan was generated much like the gross sales evaluation to calculate traits and outliers, but additionally, to visualize anomalies generated by the anomaly detection mannequin.
Right here is the plan
[
{
"step_id": 1,
"name": "Load Production Data for SOS001L12P",
"description": "Load the 'Production .csv' file from 'C:AnomalyGNNdatasetRaw DatasetTemporal DataUnit' into a pandas DataFrame. Select only the 'Date' and 'SOS001L12P' columns for analysis.",
"type": "CODE"
},
{
"step_id": 2,
"name": "Clean and Prepare Production Data",
"description": "Convert the 'Date' column to datetime objects and set it as the DataFrame index. Ensure the 'SOS001L12P' column is numeric. Handle any missing values in 'SOS001L12P' by filling with 0, assuming missing production data means zero units produced.",
"type": "CODE"
},
{
"step_id": 3,
"name": "Calculate Rolling Averages for Trend",
"description": "Compute a 7-day and a 30-day rolling average for the 'SOS001L12P' production volume. These rolling averages will help to visualize and identify short-term and long-term trends in the data.",
"type": "CODE"
},
{
"step_id": 4,
"name": "Visualize Production Volume and Trends",
"description": "Create a time series line plot showing the daily 'SOS001L12P' production volume. Overlay the 7-day and 30-day rolling averages on the same plot to visually represent the trend. Ensure proper labeling for axes and a clear title.",
"type": "CODE"
},
Z-score,
{
"step_id": 6,
"name": "Visualize Outliers and Abnormalities",
"description": "Generate a time series line plot of the 'SOS001L12P' production volume. On this plot, distinctly mark or color the data points that were identified as outliers in the previous step to highlight abnormalities in volume trends.",
"type": "CODE"
},
{
"step_id": 7,
"name": "Summarize Trend, Outliers, and Abnormalities",
"description": "Provide a textual summary of the observed overall trend (e.g., increasing, decreasing, stable, seasonal), describe the characteristics of the identified outliers (e.g., magnitude, frequency, specific dates), and comment on any notable abnormalities in the production volume patterns based on the visualizations and statistical analysis.",
"type": "TEXT"
}
]Applicable instruments had been created, executed and outputs produced. Following are the manufacturing development chart and extra importantly, the chart exhibiting a major variety of anomalies within the second half of the interval, which could be impacting gross sales.
That is famous within the report generated:
” Regardless of the dearth of Z-score outliers, a extra subtle anomaly detection methodology **recognized 17 anomalous information factors** within the SOS001L12P manufacturing.“
Full report right here
Here is an evaluation of the manufacturing temporal sign for SOS001L12P:
**1. Knowledge Overview:**
The manufacturing information for SOS001L12P was efficiently loaded and preprocessed, masking 221 days from January 1, 2023, to August 9, 2023. The 'Date' column was accurately parsed, and the 'SOS001L12P' manufacturing quantity column was confirmed to be numeric with no lacking values.
**2. Development Evaluation:**
* **Rolling Averages:** 7-day and 30-day rolling averages had been computed to easy out day by day fluctuations and spotlight underlying traits.
* The plot `SOS001L12P_production_trends.png` (saved to `C://Auto-agent/workspace/output`) visually represents the day by day manufacturing quantity alongside these rolling averages, permitting for commentary of the general temporal development.
* Wanting on the tail of the information, the 7-day and 30-day rolling averages present a common decline in manufacturing quantity in the direction of the top of the noticed interval (early August), with the 30-day common dropping from 5582.97 on August seventh to 5075.73 on August ninth, and the 7-day common dropping from 3724.71 to 2911.00 in the identical interval.
**3. Outliers (Z-score Methodology):**
* Utilizing a Z-score threshold of |Z-score| > 2.5, **no potential outliers had been recognized** within the SOS001L12P manufacturing information. This implies that whereas there could be variations, no particular person information factors considerably deviated from the imply (7480.90) by greater than 2.5 customary deviations (4729.55).
**4. Abnormalities in Quantity Tendencies:**
* Regardless of the dearth of Z-score outliers, a extra subtle anomaly detection methodology **recognized 17 anomalous information factors** within the SOS001L12P manufacturing.
* These anomalies are visualized within the plot `SOS001L12P_production_with_anomalies_plot.png` (saved to `C://Auto-agent/workspace/output`), which might spotlight particular dates the place manufacturing quantity exhibited uncommon habits in comparison with its typical sample. A big drop to 0 on August ninth, for example, is prone to be one such anomaly, suggesting a halt or extreme discount in manufacturing.Finish-to-Finish Provide Planner’s report
Now that we’ve carried out spatio-temporal evaluation on the SKU SOS001L12P, the subsequent logical step is to correlate the findings to the demand forecast mannequin to see how it’s adapting to the traits. And put together a full report for this SKU well being that the planner can take to the administration.
To do that, we set the next purpose:
“I’m a provide chain planner. I would really like you to think about the previous evaluation you probably did not too long ago together with different information sources and create a finish to finish report on SOS001L12P. I observed that the manufacturing is trending down within the second half of the length in addition to there are a lot of anomalies throughout this era. Correlate the manufacturing with the demand forecast and touch upon how our forecasting mannequin is performing through the length.“
The planner generated a plan to correlate manufacturing, gross sales and demand forecast, consider forecasting mannequin efficiency utilizing the first metric WAPE (Weighted Absolute Share Error), then put together a complete report.
Right here is the plan
[
{
"step_id": 1,
"name": "Load and Consolidate SOS001L12P Data",
"description": "Load the following datasets for SOS001L12P: n1. Production rolling averages: C:Auto-agentworkspaceoutputproduction analysis - 2production_rolling_averages_SOS001L12P.csv (columns: Date, SOS001L12P, SOS001L12P_7_day_avg, SOS001L12P_30_day_avg)n2. Production Z-Scores: C:Auto-agentworkspaceoutputproduction analysis - 2SOS001L12P_production_z_scores.csv (columns: Date, Actual_Production, Z_Score)n3. Anomaly detection results: C:AnomalyGNNresultstop_sku_SOS001L12P_predictions.csv (columns: Date, Is_Anomaly)n4. Demand vs. Production Forecast: C:GNNresultsdemand_vs_production_SOS001L12P.csv (columns: Date, Predicted_Demand, Actual_Sales, Actual_Production)nnMerge all these dataframes on the 'Date' column, ensuring 'Date' is converted to a datetime format for accurate time-series analysis.",
"type": "CODE"
},
> 2 or 3) to visually confirm the user's observation of a downward trend and anomalies in the second half of the duration. Identify the start date of the 'second half' based on the data's time range.",
"type": "CODE"
,
{
"step_id": 3,
"name": "Correlate Production, Sales, and Demand Forecast",
"description": "Generate a single time-series line plot comparing 'Actual_Production', 'Actual_Sales', and 'Predicted_Demand' for SOS001L12P. Analyze the visual correlation and note periods of significant divergence or alignment, especially during the identified 'second half' where production anomalies and downward trends were observed.",
"type": "CODE"
},
{
"step_id": 4,
"name": "Evaluate Forecasting Model Performance (WAPE)",
"description": "Calculate the Weighted Average Percentage Error (WAPE) for the 'Predicted_Demand' against 'Actual_Sales' for SOS001L12P using the consolidated data. Calculate WAPE for the entire duration and separately for the 'second half' of the duration (as identified in Step 2) to assess how the model performed during the period of observed production issues. Present the WAPE values and interpret their meaning regarding forecast accuracy.",
"type": "CODE"
},
{
"step_id": 5,
"name": "Generate Comprehensive End-to-End Report for SOS001L12P",
"description": "Provide a narrative report summarizing all findings. n1. Confirm the observed downward trend in production for SOS001L12P during the second half of the duration, citing specific dates or periods. n2. Validate the user's observation of many anomalies, providing a count or highlighting key anomaly dates. n3. Discuss the correlation between actual production, actual sales, and predicted demand, explaining how they interact, especially during the problematic second half. n4. Comment on the forecasting model's performance based on the calculated WAPE values, noting any significant differences in accuracy between the overall duration and the 'second half'. n5. Offer insights into potential reasons for observed trends or anomalies and their impact on supply chain planning.",
"type": "TEXT"
}
]The mixed evaluation produces a planner-ready evaluation, which notes in its conclusion:
“For SKU SOS001L12P, the second half of the analyzed length (April 21, 2023 – August 9, 2023) was characterised by elevated volatility and a downward development in Precise Manufacturing, accompanied by quite a few anomalies. Precise Manufacturing throughout this era was additionally decrease than Precise Gross sales. Whereas the demand forecasting mannequin successfully captured Precise Gross sales traits (WAPE of 29.70% within the second half, exhibiting an enchancment from the general 35.44%), manufacturing struggled to align with these calls for, resulting in weak correlations between manufacturing and each gross sales and predicted demand. These divergences, usually coinciding with recognized manufacturing anomalies, recommend potential provide constraints or operational disruptions impacting our means to satisfy demand persistently.“
The complete report is right here
Right here is an end-to-end report for SKU SOS001L12P, contemplating your observations and the current analyses.
---
### **Finish-to-Finish Report: SKU SOS001L12P Evaluation (2023-01-01 to 2023-08-09)**
**1. Overview**
This report synthesizes current manufacturing evaluation, anomaly detection, and demand forecasting information for SKU SOS001L12P, masking the interval from January 1, 2023, to August 9, 2023. The evaluation particularly focuses on traits and mannequin efficiency through the second half of this length, beginning roughly April 21, 2023.
**2. Manufacturing Tendencies and Anomalies**
The evaluation confirms your commentary concerning manufacturing traits within the second half of the length (from 2023-04-21).
* **Downward Development and Volatility**: Whereas the time sequence typically reveals a constructive visible correlation between Precise Manufacturing, Precise Gross sales, and Predicted Demand, Precise Manufacturing seems to be extra unstable and displays sharper drops in comparison with Precise Gross sales and Predicted Demand in sure segments of the second half.
* **Manufacturing vs. Gross sales**: Within the second half, Precise Manufacturing (common 4866 items) was notably decrease than Precise Gross sales (common 5136 items).
* **Anomalies**: A complete of 17 mixed anomaly factors (primarily based on 'Is_Anomaly=True' or |Z_Score| > 2.5) had been recognized throughout the whole dataset. A number of of those manufacturing anomalies had been detected within the second half, particularly on dates reminiscent of 2023-04-21, 2023-05-05, 2023-05-21, 2023-05-22, and 2023-05-23. These anomaly intervals usually coincide with vital divergences the place Precise Manufacturing deviates sharply from Precise Gross sales and Predicted Demand.
**3. Correlation: Manufacturing, Demand Forecast, and Gross sales**
The correlation evaluation within the second half of the length (from 2023-04-21) reveals distinct relationships:
* **Robust Gross sales-Demand Correlation**: Precise Gross sales and Predicted Demand present a robust constructive correlation of **0.85**. This implies the forecasting mannequin is mostly efficient in monitoring precise gross sales patterns.
* **Weak Manufacturing Correlation**: Precise Manufacturing's correlation with each Precise Gross sales and Predicted Demand is considerably weaker:
* Precise Manufacturing vs. Precise Gross sales: **0.35**
* Precise Manufacturing vs. Predicted Demand: **0.38**
* **Divergence**: Regardless of a common visible correlation throughout the whole interval, the second half displays noticeable intervals the place Precise Manufacturing diverges from Precise Gross sales and Predicted Demand. That is notably evident through the noticed anomalies, indicating that manufacturing output didn't persistently align with market demand or gross sales in these situations.
**4. Demand Forecasting Mannequin Efficiency**
The forecasting mannequin's efficiency was assessed utilizing the Weighted Common Share Error (WAPE):
* **Total Efficiency**: The general WAPE for SOS001L12P throughout the whole length is **35.44%**.
* **Second Half Efficiency**: For the 'second half' of the length (from 2023-04-21), the WAPE is **29.70%**.
* **Interpretation**: The WAPE within the second half is decrease than the general WAPE. This means that, regardless of the noticed manufacturing points and anomalies throughout this era, the demand forecasting mannequin's accuracy truly **improved** within the second half. The mannequin's Predicted Demand persistently tracks Precise Gross sales intently, suggesting it's fairly efficient in capturing demand patterns for planning functions.
**Conclusion**
For SKU SOS001L12P, the second half of the analyzed length (April 21, 2023 - August 9, 2023) was characterised by elevated volatility and a downward development in Precise Manufacturing, accompanied by quite a few anomalies. Precise Manufacturing throughout this era was additionally decrease than Precise Gross sales. Whereas the demand forecasting mannequin successfully captured Precise Gross sales traits (WAPE of 29.70% within the second half, exhibiting an enchancment from the general 35.44%), manufacturing struggled to align with these calls for, resulting in weak correlations between manufacturing and each gross sales and predicted demand. These divergences, usually coinciding with recognized manufacturing anomalies, recommend potential provide constraints or operational disruptions impacting our means to satisfy demand persistently.This can be a production-grade report consisting of information, figures and KPIs to assist the interpretations derived.
And all of this was achieved with a very versatile agentic framework that may adapt to any explainability purpose desired.
Growing and testing this utility, together with producing the above experiences and outcomes price lower than $1 in token utilization.
Conclusion
The way forward for agentic techniques shouldn’t be bigger software catalogs or extra elaborate orchestration frameworks. It’s brokers that may determine what must exist within the first place.
When planning, coding, and execution are handled as reasoning steps, instruments cease being static belongings and develop into disposable artifacts—created, used, and discarded as the issue calls for. This shifts agent design away from immediate engineering attempting to bind the brokers and instruments to the orchestration framework towards software program technology as a core functionality.
The plan–code–execute structure demonstrates that production-grade, evidence-based evaluation can emerge from small, clear brokers with minimal glue code and minimal price. As LLMs proceed to enhance, essentially the most priceless brokers will likely be these that may autonomously flip intent into working software program—with out being advised prematurely find out how to do it.
In that sense, auto-planning and auto-coding brokers are usually not only a comfort. They’re a pure evolution of agentic techniques.
For extra on the GNN mannequin for demand forecasting referred on this article, learn my article right here.
Join with me and share your feedback at www.linkedin.com/in/partha-sarkar-lets-talk-AI
Reference
SupplyGraph: A Benchmark Dataset for Provide Chain Planning utilizing Graph Neural Networks : Authors: Azmine Toushik Wasi, MD Shafikul Islam, Adipto Raihan Akib
Photos used on this article are generated utilizing Google Gemini. Charts and underlying code created by me.
