diff --git a/.github/test_files/config_rasa_demo.yml b/.github/test_files/config_rasa_demo.yml
new file mode 100644
index 0000000..09f976d
--- /dev/null
+++ b/.github/test_files/config_rasa_demo.yml
@@ -0,0 +1,49 @@
+language: en
+pipeline:
+- name: WhitespaceTokenizer
+- name: RegexFeaturizer
+- name: LexicalSyntacticFeaturizer
+- name: CountVectorsFeaturizer
+  OOV_token: oov
+  token_pattern: (?u)\b\w+\b
+- name: CountVectorsFeaturizer
+  analyzer: char_wb
+  min_ngram: 1
+  max_ngram: 4
+- name: DIETClassifier
+  epochs: 5
+  ranking_length: 5
+- name: DucklingEntityExtractor
+  url: http://localhost:8000
+  dimensions:
+  - email
+  - number
+  - amount-of-money
+- name: EntitySynonymMapper
+- name: ResponseSelector
+  retrieval_intent: out_of_scope
+  scale_loss: false
+  epochs: 5
+- name: ResponseSelector
+  retrieval_intent: faq
+  scale_loss: false
+  epochs: 5
+- name: ResponseSelector
+  retrieval_intent: chitchat
+  scale_loss: false
+  epochs: 5
+- name: FallbackClassifier
+  threshold: 0.7
+policies:
+- name: TEDPolicy
+  max_history: 5
+  epochs: 10
+  batch_size:
+  - 32
+  - 64
+- max_history: 6
+  name: AugmentedMemoizationPolicy
+- name: RulePolicy
+  core_fallback_threshold: 0.3
+  core_fallback_action_name: "action_default_fallback"
+  enable_fallback_prediction: True
diff --git a/.github/test_files/config_rasa_x_demo.yml b/.github/test_files/config_rasa_x_demo.yml
new file mode 100644
index 0000000..eba35bc
--- /dev/null
+++ b/.github/test_files/config_rasa_x_demo.yml
@@ -0,0 +1,36 @@
+# Configuration for Rasa NLU.
+# https://rasa.com/docs/rasa/nlu/components/
+language: en
+
+pipeline:
+# # No configuration for the NLU pipeline was provided. The following default pipeline was used to train your model.
+# # If you'd like to customize it, uncomment and adjust the pipeline.
+# # See https://rasa.com/docs/rasa/tuning-your-model for more information.
+   - name: WhitespaceTokenizer
+   - name: RegexFeaturizer
+   - name: LexicalSyntacticFeaturizer
+   - name: CountVectorsFeaturizer
+   - name: CountVectorsFeaturizer
+     analyzer: char_wb
+     min_ngram: 1
+     max_ngram: 4
+   - name: DIETClassifier
+     epochs: 5
+   - name: EntitySynonymMapper
+   - name: ResponseSelector
+     epochs: 5
+   - name: FallbackClassifier
+     threshold: 0.3
+     ambiguity_threshold: 0.1
+
+# Configuration for Rasa Core.
+# https://rasa.com/docs/rasa/core/policies/
+policies:
+# # No configuration for policies was provided. The following default policies were used to train your model.
+# # If you'd like to customize them, uncomment and adjust the policies.
+# # See https://rasa.com/docs/rasa/policies for more information.
+   - name: MemoizationPolicy
+   - name: TEDPolicy
+     max_history: 5
+     epochs: 5
+   - name: RulePolicy
diff --git a/.github/test_files/report_to_compare.json b/.github/test_files/report_to_compare.json
new file mode 100644
index 0000000..2120e73
--- /dev/null
+++ b/.github/test_files/report_to_compare.json
@@ -0,0 +1,76 @@
+{
+    "default": {
+      "Rasa Demo - NLU": {
+        "entity_prediction": {
+          "macro_avg": {
+            "f1-score": 0.2378897757789324,
+            "precision": 0.7177262292340671,
+            "recall": 0.630076550644482,
+            "support": 1451
+          },
+          "micro_avg": {
+            "f1-score": 0.557902566009669,
+            "precision": 0.8231017770597738,
+            "recall": 0.7022742935906272,
+            "support": 1451
+          },
+          "weighted_avg": {
+            "f1-score": 0.6357964876470346,
+            "precision": 0.8108170836964997,
+            "recall": 0.7022742935906272,
+            "support": 1451
+          }
+        },
+        "intent_classification": {
+          "macro_avg": {
+            "f1-score": 0.78507804141552495,
+            "precision": 0.6249297611496761,
+            "recall": 0.22805363055392253,
+            "support": 4950
+          },
+          "micro_avg": {
+            "f1-score": 0.5612119662490412,
+            "precision": 0.9004178272980501,
+            "recall": 0.5224242424242425,
+            "support": 4950
+          },
+          "weighted_avg": {
+            "f1-score": 0.5845322720747205,
+            "precision": 0.8560612861880288,
+            "recall": 0.5224242424242425,
+            "support": 4950
+          }
+        },
+        "response_selection": {
+          "accuracy": 0.657748611812216,
+          "macro_avg": {
+            "f1-score": 0.4511394032023792,
+            "precision": 0.541429844814056,
+            "recall": 0.47505801532174247,
+            "support": 1981
+          },
+          "weighted_avg": {
+            "f1-score": 0.6172915897390211,
+            "precision": 0.6469121522031172,
+            "recall": 0.657748611812216,
+            "support": 1981
+          }
+        },
+        "story_prediction": {
+          "accuracy": 0.51005291005291,
+          "macro_avg": {
+            "f1-score": 0.5328584995251662,
+            "precision": 0.8363425925925926,
+            "recall": 0.8482142857142857,
+            "support": 189
+          },
+          "weighted_avg": {
+            "f1-score": 0.5091936869714647,
+            "precision": 0.9186067019400354,
+            "recall": 0.91005291005291,
+            "support": 189
+          }
+        }
+      }
+    }
+  }
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
new file mode 100644
index 0000000..b878c34
--- /dev/null
+++ b/.github/workflows/ci.yml
@@ -0,0 +1,161 @@
+name: Continuous Integration
+on:
+  push:
+    branches:
+    - master
+    tags:
+    - '*'
+  pull_request:
+
+env:
+  RASA_X_DEMO_VERSION: '0.33.0'
+  RASA_IMAGE_TAG: '2.1.3-full'
+
+jobs:
+  test_action_init:
+    runs-on: ubuntu-latest
+    name: Test GH action - rasa init
+
+    steps:
+    - uses: actions/checkout@v2
+
+    - name: Create a new project
+      run: |
+        mkdir rasa-project
+        chmod 777 rasa-project
+        docker run --rm -v ${{ github.workspace }}/rasa-project:/rasa-project rasa/rasa:${{ env.RASA_IMAGE_TAG }} init --no-prompt --init-dir /rasa-project
+
+    - name: Download Rasa X model
+      run: |
+        wget https://github.com/RasaHQ/rasa-x-demo/blob/0.33.0/models/model.tar.gz?raw=true \
+          -O test_model.tar.gz
+
+    - id: action
+      name: Rasa Test & Train - rasa init
+      uses: ./
+      with:
+        rasa_version: ${{ env.RASA_IMAGE_TAG }}
+        workspace: ${{ github.workspace }}/rasa-project
+        test_type: all
+        github_token: ${{ secrets.GITHUB_TOKEN }}
+        configuration_name: 'rasa-init'
+        publish_summary: 'false'
+
+    - name: Checkout output
+      run: |
+        echo ${{ fromJSON(steps.action.outputs.report).default['rasa-init'].intent_classification.accuracy }}
+
+    - name: Checkout output - if
+      if: fromJSON(steps.action.outputs.report).default['rasa-init'].intent_classification.accuracy <= 0.5
+      run: |
+        echo "::error::${{ fromJSON(steps.action.outputs.report).default['rasa-init'].intent_classification.accuracy }}"
+        exit 1
+
+  # Test the GH action on Rasa X Demo data
+  # https://github.com/RasaHQ/rasa-x-demo
+  test_action_x:
+    runs-on: ubuntu-latest
+    name: Test GH action - Rasa X Demo
+
+    steps:
+    - uses: actions/checkout@v2
+
+    - name: Checkout Rasa X Demo
+      uses: actions/checkout@v2
+      with:
+        repository: RasaHQ/rasa-x-demo
+        path: rasa-x-demo
+        ref: ${{ env.RASA_X_DEMO_VERSION }}
+
+    - name: Download Rasa X model
+      run: |
+        wget https://github.com/RasaHQ/rasa-x-demo/blob/0.33.0/models/model.tar.gz?raw=true \
+          -O test_model.tar.gz
+
+    - id: action
+      name: Rasa Test & Train - Rasa X Demo
+      uses: ./
+      with:
+        rasa_version: '2.0.0-full'
+        workspace: ${{ github.workspace }}/rasa-x-demo
+        test_type: all
+        github_token: ${{ secrets.GITHUB_TOKEN }}
+        model: github_workspace/test_model.tar.gz
+        configuration: github_workspace/test_files/config_rasa_x_demo.yml
+        configuration_name: 'Rasa X Demo'
+        publish_summary: 'false'
+
+    - name: Checkout output
+      run: |
+        echo ${{ fromJSON(steps.action.outputs.report).default['Rasa X Demo'].intent_classification.accuracy }}
+
+    - name: Checkout output - if
+      if: fromJSON(steps.action.outputs.report).default['Rasa X Demo'].intent_classification.accuracy <= 0.5
+      run: |
+        echo "::error::${{ fromJSON(steps.action.outputs.report).default['Rasa X Demo'].intent_classification.accuracy }}"
+        exit 1
+
+  # Test the GH action on Rasa Demo data
+  # https://github.com/RasaHQ/rasa-x-demo
+  # Test type: NLU
+  test_action:
+    runs-on: ubuntu-latest
+    name: Test GH action - Rasa Demo - NLU
+    needs: [test_action_x]
+
+    steps:
+    - uses: actions/checkout@v2
+
+    - name: Checkout Rasa Demo
+      uses: actions/checkout@v2
+      with:
+        repository: RasaHQ/rasa-demo
+        path: rasa-demo
+
+    - name: Use test configuration
+      run: |
+        cp .github/test_files/config_rasa_demo.yml rasa-demo/config.yml
+
+    - id: action
+      name: Rasa Test & Train - Rasa Demo
+      uses: ./
+      with:
+        rasa_version: '2.0.0-full'
+        configuration_name: 'Rasa Demo - NLU'
+        workspace: ${{ github.workspace }}/rasa-demo
+        test_type: nlu
+        github_token: ${{ secrets.GITHUB_TOKEN }}
+        test_args: '-f 2'
+        # Use cross validation
+        cross_validation: 'true'
+        # Compare results to the test report
+        compare_report: .github/test_files/report_to_compare.json
+
+  # Test the GH action on Rasa Demo data
+  # https://github.com/RasaHQ/rasa-x-demo
+  # Test type: ALL
+  test_action_all:
+    runs-on: ubuntu-latest
+    name: Test GH action - Rasa Demo - ALL
+
+    steps:
+    - uses: actions/checkout@v2
+
+    - name: Checkout Rasa Demo
+      uses: actions/checkout@v2
+      with:
+        repository: RasaHQ/rasa-demo
+        path: rasa-demo
+
+    - name: Use test configuration
+      run: |
+        cp .github/test_files/config_rasa_demo.yml rasa-demo/config.yml
+
+    - id: action
+      name: Rasa Test & Train - Rasa Demo
+      uses: ./
+      with:
+        rasa_version: '2.0.0-full'
+        configuration_name: 'Rasa Demo'
+        workspace: ${{ github.workspace }}/rasa-demo
+        github_token: ${{ secrets.GITHUB_TOKEN }}
diff --git a/.github/workflows/ci_summary.yml b/.github/workflows/ci_summary.yml
new file mode 100644
index 0000000..a773111
--- /dev/null
+++ b/.github/workflows/ci_summary.yml
@@ -0,0 +1,79 @@
+name: Continuous Integration - Test summary
+on:
+  push:
+    branches:
+    - master
+    tags:
+    - '*'
+  pull_request:
+
+jobs:
+  test_action_1:
+    runs-on: ubuntu-latest
+    name: Test GH action - Rasa Demo - cfg 1
+
+    steps:
+    - uses: actions/checkout@v2
+
+    - name: Checkout Rasa Demo
+      uses: actions/checkout@v2
+      with:
+        repository: RasaHQ/rasa-demo
+        path: rasa-demo
+
+    - name: Use test configuration
+      run: |
+        cp .github/test_files/config_rasa_demo.yml rasa-demo/config.yml
+
+    - id: action
+      name: Rasa Test & Train - Rasa Demo
+      uses: ./
+      with:
+        rasa_version: '2.0.0-full'
+        configuration_name: 'Rasa Demo - Configuration 1'
+        workspace: ${{ github.workspace }}/rasa-demo
+        publish_summary: 'false'
+
+    - uses: actions/upload-artifact@v2
+      with:
+        name: rasa-demo-cfg-1
+        path: rasa-demo/reports
+
+  test_action_2:
+    runs-on: ubuntu-latest
+    name: Test GH action - Rasa Demo - cfg 2
+    needs: [test_action_1]
+
+    steps:
+    - uses: actions/checkout@v2
+
+    - name: Checkout Rasa Demo
+      uses: actions/checkout@v2
+      with:
+        repository: RasaHQ/rasa-demo
+        path: rasa-demo
+
+    - uses: actions/download-artifact@v2
+      with:
+        name: rasa-demo-cfg-1
+        path: rasa-demo/reports
+
+    - name: Use test configuration
+      run: |
+        cp .github/test_files/config_rasa_demo.yml rasa-demo/config.yml
+
+    - id: action
+      name: Rasa Test & Train - Rasa Demo
+      uses: ./
+      with:
+        rasa_version: '2.0.0-full'
+        configuration_name: 'Rasa Demo - Configuration 2'
+        workspace: ${{ github.workspace }}/rasa-demo
+        github_token: ${{ secrets.GITHUB_TOKEN }}
+
+    - uses: actions/upload-artifact@v2
+      with:
+        name: rasa-demo-cfg-2
+        path: |
+          rasa-demo/results
+          rasa-demo/reports
diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml
new file mode 100644
index 0000000..7140dbb
--- /dev/null
+++ b/.github/workflows/release.yml
@@ -0,0 +1,41 @@
+name: Release
+on:
+  push:
+    tags:
+    - '*'
+
+jobs:
+  release:
+    name: Create Release
+    runs-on: ubuntu-latest
+    if: startsWith(github.event.ref, 'refs/tags')
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v2
+        with:
+          fetch-depth: 0
+
+      - name: Get release version
+        id: release_version
+        run: |
+          sed -e 's/_version_/'${GITHUB_REF#refs/tags/}'/g' package.json.tmp > package.json
+          echo "::set-output name=version::${GITHUB_REF#refs/tags/}"
+
+      - name: Changelog
+        uses: scottbrenner/generate-changelog-action@1.0.1
+        id: changelog
+        env:
+          REPO: ${{ github.repository }}
+
+      - name: Create Release
+        id: create_release
+        uses: actions/create-release@v1
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        with:
+          tag_name: ${{ steps.release_version.outputs.version }}
+          release_name: ${{ steps.release_version.outputs.version }}
+          body: |
+            ${{ steps.changelog.outputs.changelog }}
+          draft: false
+          prerelease: false
diff --git a/.gitignore b/.gitignore
new file mode 100644
index 0000000..b3f8862
--- /dev/null
+++ b/.gitignore
@@ -0,0 +1,3 @@
+rasa-demo
+results.json
+report_master.json
diff --git a/README.md b/README.md
index db1649b..7fcaf39 100644
--- a/README.md
+++ b/README.md
@@ -1 +1,229 @@
-# rasa-train-test-gha
\ No newline at end of file
+# Rasa Train-Test Model GitHub Action
+
+The GitHub action to run easily `rasa train` and `rasa test` in the CIs.
+
+## Input Arguments
+
+In order to pass the input parameters to the GH action, you have to use the [`with`](https://docs.github.com/en/actions/reference/workflow-syntax-for-github-actions#jobsjob_idstepswith) argument in a step that uses the GH action, e.g.:
+
+```yaml
+jobs:
+  my_first_job:
+    name: My first job
+    runs-on: ubuntu-latest
+    steps:
+      # Checks-out GitHub repository
+      # more: https://github.com/actions/checkout
+      - uses: actions/checkout@v2
+
+      # Run rasa train and rasa test
+      - name: My first step
+        uses: RasaHQ/rasa-train-test-gha@main
+        with:
+          test_type: nlu
+```
+
+|        Input         |                                  Description                                  |          Default          |
+| -------------------- | ----------------------------------------------------------------------------- | ------------------------- |
+| `rasa_version`       | The Rasa Open Source version used to run test and train                       | `latest-full`             |
+| `rasa_image`         | Custom Rasa Docker image. Useful if you use Rasa with custom Python modules.  | `none`                    |
+| `rasa_train`         | Run `rasa train`                                                              | `true`                    |
+| `rasa_test`          | Run `rasa test`                                                               | `true`                    |
+| `data_validate`      | Validates domain and data files to check for possible mistakes                | `true`                    |
+| `workspace`          | The root directory containing your Rasa Open Source project                   | `${{ github.workspace }}` |
+| `train_args`         | Additional arguments passed to the `rasa train` command                       | `none`                    |
+| `test_args`          | Additional arguments passed to the `rasa test` command                        | `none`                    |
+| `test_type`          | The types of tests to run (available types: `core`/`nlu`/`all`)               | `all`                     |
+| `publish_summary`    | Publish tests summary as a PR comment                                         | `true`                    |
+| `github_token`       | GitHub Token - required to add a comment with summary                         | `none`                    |
+| `configuration`      | Model configuration file                                                      | `config.yml`              |
+| `model`              | Path to a file with a model. Use existing model instead of training a new one | `none`                    |
+| `cross_validation`   | Switch on cross validation mode. Any provided model will be ignored           | `false`                   |
+| `configuration_name` | Configuration name used in summary. If not provided a file name is used       | `none`                    |
+| `data_name`          | Data name used in summary. If not provided a directory name is used           | `default`                 |
+| `compare_report`     | Path to a report that will be used to compare results                         | `none`                    |
+| `result_directory`   | Directory name where results are stored in                                    | `results`                 |
+| `report_directory`   | Directory name where reports are stored in                                    | `reports`                 |
+
+## Outputs
+
+The list of available output variables:
+
+|  Output  |      Description      |
+| -------- | --------------------- |
+| `report` | Return report as JSON |
+
+## Report
+
+The GH action generates two reports, a report with a summary of run tests, the report is available as JSON and CSV file. The example of a report generated by the GH action can be found [here](examples/).
+
+## Example Usage
+
+In the example below, we are using [the Rasa Demo data](https://github.com/RasaHQ/rasa-demo).
+
+```yaml
+jobs:
+    train_and_test:
+        # ...
+        steps:
+            # ...
+            - name: Train and Test Rasa Demo
+              uses: RasaHQ/rasa-train-test-gha@main
+              with:
+                # List of available tags: https://hub.docker.com/r/rasa/rasa/tags
+                rasa_version: '2.0.0-full'
+                # In order to add a PR comment with summary
+                # a GH Token has to be pass to the GH action
+                github_token: ${{ secrets.GITHUB_TOKEN }}
+            # ...
+```
+
+The GitHub action by default adds a PR comment with summary (the summary comment can be disabled by setting the `publish_summary` input argument to `false`):
+
+![Summary](_img/img_1.png)
+
+### Use JSON report in GH workflow
+
+The GH action returns a JSON report as an output. The following example shows how to use the output in a GH workflow.
+
+```yaml
+jobs:
+    train_and_test:
+        # ...
+        steps:
+            # ...
+            - name: Train and Test Rasa Demo
+              id: action
+              uses: RasaHQ/rasa-train-test-gha@main
+              with:
+                # List of available tags: https://hub.docker.com/r/rasa/rasa/tags
+                rasa_version: '2.0.0-full'
+                # In order to add a PR comment with summary
+                # a GH Token has to be pass to the GH action
+                github_token: ${{ secrets.GITHUB_TOKEN }}
+
+            # We have to convert the output to JSON by using fromJSON built-in function
+            # more: https://docs.github.com/en/free-pro-team@latest/actions/reference/context-and-expression-syntax-for-github-actions#fromjson
+            # syntax: fromJSON(steps.action.outputs.report).<data_name>[<configuration_name|configuration>]
+            # example: fromJSON(steps.action.outputs.report).default['config.yml']
+            - name: Check output
+              if: fromJSON(steps.action.outputs.report).default['config.yml'].intent_classification.accuracy >= 0.8
+              run: |
+                echo "I'm doing extra work..."
+                echo ${{ fromJSON(steps.action.outputs.report).default['config.yml'].intent_classification.accuracy }}
+```
+
+## Advanced examples
+
+### Compare results with the base report
+
+It possible to compare results to the other report, the feature is useful for example to see that a model is better than before changes. The difference against the report that we compare to is included in brackets.
+
+```yaml
+jobs:
+    train_and_test:
+        # ...
+        steps:
+            # ...
+            - name: Train and Test Rasa Demo
+              uses: RasaHQ/rasa-train-test-gha@main
+              with:
+                # List of available tags: https://hub.docker.com/r/rasa/rasa/tags
+                rasa_version: '2.0.0-full'
+                # In order to add a PR comment with summary
+                # a GH Token has to be pass to the GH action
+                github_token: ${{ secrets.GITHUB_TOKEN }}
+                # A path to the report that we want to compare to
+                compare_report: 'report_to_compare.json'
+                test_type: 'nlu'
+            # ...
+```
+
+![Compare](_img/img_3.png)
+
+### Use the existing model
+
+It's possible to use the existing model instead of training a new one.
+
+```yaml
+jobs:
+    test:
+        # ...
+        steps:
+            # ...
+            - name: Download Rasa X model
+              run: |
+                wget https://github.com/RasaHQ/rasa-x-demo/blob/0.33.0/models/model.tar.gz?raw=true \
+                  -O test_model.tar.gz
+
+            - name: Train and Test Rasa Demo
+              uses: RasaHQ/rasa-train-test-gha@main
+              with:
+                # List of available tags: https://hub.docker.com/r/rasa/rasa/tags
+                rasa_version: '2.0.0-full'
+                # In order to add a PR comment with summary
+                # a GH Token has to be pass to the GH action
+                github_token: ${{ secrets.GITHUB_TOKEN }}
+                # If a file with the model is provided, training is disabled automatically
+                model: test_model.tar.gz
+            # ...
+```
+
+### Cross-validate NLU model
+
+```yaml
+jobs:
+    train_and_test:
+        # ...
+        steps:
+            # ...
+            - name: Train and Test Rasa Demo
+              uses: RasaHQ/rasa-train-test-gha@main
+              with:
+                # List of available tags: https://hub.docker.com/r/rasa/rasa/tags
+                rasa_version: '2.0.0-full'
+                # In order to add a PR comment with summary
+                # a GH Token has to be pass to the GH action
+                github_token: ${{ secrets.GITHUB_TOKEN }}
+                # Switch on cross validation mode. Any provided model will be ignored
+                cross_validation: 'true'
+                # Number of cross validation folds (cross validation only)
+                # All available arguments: rasa test nlu --help
+                test_args: '--folds 3'
+                test_type: 'nlu'
+            # ...
+```
+
+In a case where the cross-validation mode is enabled, a summary published as a PR comment includes `Intent Cross-Validation Results` and `Entity Cross-Validation Results`, e.g.
+
+![Intent Cross-Validation Results](_img/img_4.png)
+
+![Entity Cross-Validation Results](_img/img_5.png)
+
+### Upload results and reports as artifacts
+
+```yaml
+jobs:
+    train_and_test:
+        # ...
+        steps:
+            # ...
+            - name: Train and Test Rasa Demo
+              id: action
+              uses: RasaHQ/rasa-train-test-gha@main
+              with:
+                # List of available tags: https://hub.docker.com/r/rasa/rasa/tags
+                rasa_version: '2.0.0-full'
+                # In order to add a PR comment with summary
+                # a GH Token has to be pass to the GH action
+                github_token: ${{ secrets.GITHUB_TOKEN }}
+
+            # The 'actions/upload-artifact' action to upload files
+            # More: https://github.com/actions/upload-artifact
+            - uses: actions/upload-artifact@v2
+              with:
+                name: rasa-demo-cfg
+                path: |
+                  results
+                  reports
+```
diff --git a/_img/img_1.png b/_img/img_1.png
new file mode 100644
index 0000000..567b146
Binary files /dev/null and b/_img/img_1.png differ
diff --git a/_img/img_2.png b/_img/img_2.png
new file mode 100644
index 0000000..d74535c
Binary files /dev/null and b/_img/img_2.png differ
diff --git a/_img/img_3.png b/_img/img_3.png
new file mode 100644
index 0000000..e69d7f7
Binary files /dev/null and b/_img/img_3.png differ
diff --git a/_img/img_4.png b/_img/img_4.png
new file mode 100644
index 0000000..1d55a18
Binary files /dev/null and b/_img/img_4.png differ
diff --git a/_img/img_5.png b/_img/img_5.png
new file mode 100644
index 0000000..cc029ea
Binary files /dev/null and b/_img/img_5.png differ
diff --git a/action.yml b/action.yml
new file mode 100644
index 0000000..3484f93
--- /dev/null
+++ b/action.yml
@@ -0,0 +1,267 @@
+name: 'Rasa Train-Test Model GitHub Action'
+description: 'The GitHub action to run the rasa train and rasa test commands'
+inputs:
+  rasa_version:
+    description: 'The Rasa version used to run test and train'
+    required: true
+    default: 'latest-full'
+  rasa_image:
+    description: 'Custom Docker image'
+    required: false
+    default: ''
+  rasa_train:
+    description: 'Run rasa train'
+    required: true
+    default: 'true'
+  rasa_test:
+    description: 'Run rasa test'
+    required: true
+    default: 'true'
+  data_validate:
+    description: 'Validates domain and data files to check for possible mistakes'
+    required: true
+    default: 'true'
+  workspace:
+    description: 'The root directory'
+    required: false
+    default: '${{ github.workspace }}'
+  train_args:
+    description: 'Additional arguments passed to the rasa train command'
+    required: false
+    default: ''
+  test_args:
+    description: 'Additional arguments passed to the rasa test command'
+    required: false
+    default: ''
+  test_type:
+    description: 'The types of tests to run (available types: core/nlu/all)'
+    required: true
+    default: 'all'
+  publish_summary:
+    description: 'Publish tests summary as a PR comment'
+    required: false
+    default: 'true'
+  github_token:
+    description: 'GitHub Token - required to add a comment with summary'
+    required: false
+    default: ''
+  configuration:
+    description: 'Model configuration file'
+    required: true
+    default: 'config.yml'
+  model:
+    description: 'Path to a file with a model. Use existing model instead of training a new one'
+    required: false
+    default: ''
+  cross_validation:
+    description: 'Switch on cross validation mode. Any provided model will be ignored'
+    required: false
+    default: 'false'
+  configuration_name:
+    description: 'Configuration name used in summary. If not provided a file name is used'
+    required: false
+    default: ''
+  data_name:
+    description: 'Data name used in summary. If not provided a directory name is used'
+    required: false
+    default: 'default'
+  compare_report:
+    description: 'Path to a report that will be used to compare results'
+    required: false
+    default: ''
+  result_directory:
+    description: 'Directory name where results are stored in'
+    required: false
+    default: 'results'
+  report_directory:
+    description: 'Directory name where reports are stored in'
+    required: false
+    default: 'reports'
+outputs:
+  report:
+    description: "Return report as JSON"
+    value: ${{ steps.report_output.outputs.report }}
+
+branding:
+  icon: 'layers'
+  color: 'green'
+
+runs:
+  using: "composite"
+  steps:
+    - name: Set environment
+      shell: bash
+      run: |-
+        mkdir -p ${{ inputs.workspace }}/models
+        chmod 777 ${{ inputs.workspace }}/models
+        mkdir -p ${{ inputs.workspace }}/${{ inputs.result_directory }}
+        chmod 777 ${{ inputs.workspace }}/${{ inputs.result_directory }}
+
+        case "${{ inputs.test_type }}" in
+          all)
+            echo TEST_TYPE="" >> $GITHUB_ENV
+            ;;
+          nlu)
+            echo TEST_TYPE=nlu >> $GITHUB_ENV
+            echo DEFAULT_ARGS="--config /app/${{ inputs.configuration }}" >> $GITHUB_ENV
+            ;;
+          core)
+            echo TEST_TYPE=core >> $GITHUB_ENV
+            ;;
+          *)
+            echo "::error::Unknown test type: ${{ inputs.test_type }} (use: core/nlu/all)"
+            exit 1
+        esac
+
+        # Set a model path
+        if [[ -n "${{ inputs.model }}" ]];then
+          echo MODEL_ARGS="--model /app/${{ inputs.model }}" >> $GITHUB_ENV
+        fi
+
+        # Set cross validation
+        if [[ "${{ inputs.cross_validation }}" == "true" ]];then
+          echo CROSS_VALIDATION_ARGS="--cross-validation" >> $GITHUB_ENV
+        fi
+
+        # Set docker args
+        echo DOCKER_ARGS="--rm --tmpfs /.config -e MPLCONFIGDIR=/tmp/.mplconfigdir \
+          -e CI=${CI} -e RASA_TELEMETRY_ENABLED=${{ env.RASA_TELEMETRY_ENABLED }} -v ${{ github.workspace }}:/app/github_workspace \
+          -v ${{ inputs.workspace }}:/app -v ${{ inputs.workspace }}/models:/app/models \
+          -v ${{ inputs.workspace }}/${{ inputs.result_directory }}:/app/results" >> $GITHUB_ENV
+
+        # Set environment variables for generating a report
+        echo "SUMMARY_FILE=${{ inputs.workspace }}/reports/report.json" >> $GITHUB_ENV
+
+        if [[ -n "${{ inputs.configuration_name }}" ]]; then
+          echo "CONFIG=${{ inputs.configuration_name }}" >> $GITHUB_ENV
+        else
+          echo "CONFIG=${{ inputs.configuration }}" >> $GITHUB_ENV
+        fi
+
+        echo "DATASET_NAME=${{ inputs.data_name }}" >> $GITHUB_ENV
+        echo "RESULT_DIR=${{ inputs.workspace }}/${{ inputs.result_directory }}" >> $GITHUB_ENV
+
+    - name: Set the Rasa version and image
+      shell: bash
+      run: |
+        # Set custom rasa image
+        if [[ -n "${{ inputs.rasa_image }}" ]]; then
+          echo "RASA_IMAGE=${{ inputs.rasa_image }}" >> $GITHUB_ENV
+          # Don't continue if a custom rasa image was used
+          exit 0
+        fi
+
+        DOCKERHUB_TAGS_URL="https://registry.hub.docker.com/v2/repositories/rasa/rasa/tags?page_size=10000"
+        # Get the latest version
+        if [[ "${{ inputs.rasa_version }}" == "latest-full" ]]; then
+          LATEST_RASA_VERSION=$(curl -s ${DOCKERHUB_TAGS_URL} | jq -r '.results[].name' | grep -E 'full' | grep -vE 'latest' | sort -Vr | head -n1)
+          RASA_VERSION=${LATEST_RASA_VERSION}
+        else
+          # Validate Rasa version
+          CHECK_VERSION=$(curl --silent ${DOCKERHUB_TAGS_URL} | jq -r '.results[] | select(.name=="${{ inputs.rasa_version }}") | .name')
+          if [[ "$CHECK_VERSION" != "${{ inputs.rasa_version }}" ]]; then
+            echo "::error::Rasa in ${{ inputs.rasa_version }} version doesn't exist. Check if the given Rasa version is valid, https://hub.docker.com/r/rasa/rasa/tags" && exit 1
+          fi
+          RASA_VERSION=${{ inputs.rasa_version }}
+        fi
+
+        echo "RASA_IMAGE=rasa/rasa:${RASA_VERSION}" >> $GITHUB_ENV
+
+    - name: Data Validation
+      shell: bash
+      run: |
+        if [[ "${{ inputs.data_validate }}" == "true" ]]; then
+          echo "Run data validation"
+          echo
+          echo docker run ${{ env.DOCKER_ARGS }} ${{ env.RASA_IMAGE }} data validate
+          docker run ${{ env.DOCKER_ARGS }} ${{ env.RASA_IMAGE }} data validate
+        else
+          echo "::warning::Data validation is disabled. To turn on data validation set the data_validate parameter to 'true'."
+        fi
+
+    - name: Run the rasa train command
+      shell: bash
+      run: |
+        if [[ -n "${{ inputs.model }}" ]]; then
+          echo
+          echo "Use model ${{ inputs.model }}. Skipping training"
+          echo
+        elif [[ "${{ inputs.rasa_train }}" == "true" ]]; then
+          echo "Run the rasa train"
+          echo ""
+          docker run ${{ env.DOCKER_ARGS }} ${{ env.RASA_IMAGE }} train ${{ env.TEST_TYPE }} ${{ env.DEFAULT_ARGS }} ${{ inputs.train_args }}
+        else
+          echo "::warning::rasa train is disabled. To turn on the rasa train set the rasa_train parameter to 'true'."
+        fi
+
+    - name: Run the test command
+      shell: bash
+      run: |
+        if [[ "${{ inputs.rasa_test }}" == "true" ]]; then
+          echo "Run the rasa test"
+          echo ""
+          docker run ${{ env.DOCKER_ARGS }} ${{ env.RASA_IMAGE }} test ${{ env.TEST_TYPE }} ${{ env.MODEL_ARGS }} ${{ env.CROSS_VALIDATION_ARGS }} ${{ env.DEFAULT_ARGS }} ${{ inputs.test_args }}
+        else
+          echo "::warning::rasa test is disabled. To turn on the rasa train set the rasa_test parameter to 'true'."
+        fi
+
+    - name: Generate a report
+      id: report_output
+      shell: bash
+      run: |
+        mkdir -p ${{ inputs.workspace }}/${{ inputs.report_directory }}
+        python3 ${{ github.action_path }}/scripts/generate_results_report.py
+
+        # Copy a report from the current one if the report to compare to wasn't provided
+        cp ${{ inputs.workspace }}/${{ inputs.report_directory }}/report.json ${{ inputs.workspace }}/${{ inputs.report_directory }}/report_master.json
+
+        if [[ -n "${{ inputs.compare_report }}" && -e "${{ inputs.compare_report }}" ]]; then
+          cp ${{ inputs.compare_report }} ${{ inputs.workspace }}/${{ inputs.report_directory }}/report_master.json
+        elif [[ -n "${{ inputs.compare_report }}" && ! -e "${{ inputs.compare_report }}" ]]; then
+          echo "::warning::The report ${{ inputs.compare_report }} to compare doesn't exist."
+        fi
+
+        echo "::set-output name=report::$(docker run --rm -v ${{ github.action_path }}/templates:/templates \
+          -v ${{ inputs.workspace }}/${{ inputs.report_directory }}/report.json:/report.json hairyhenderson/gomplate -d report=/report.json \
+          -f /templates/report_to_json.tmpl)"
+
+        # Convert JSON to CSV
+        python3 ${{ github.action_path }}/scripts/json_to_csv.py ${{ inputs.workspace }}/${{ inputs.report_directory }}/report.json ${{ inputs.workspace }}/${{ inputs.report_directory }}/report.csv
+
+    - name: Publish summary as a comment in a PR
+      shell: bash
+      run: |
+        if [[ -z "${{ inputs.github_token }}" && "${{ inputs.publish_summary }}" == "true" && "${{ github.event_name }}" == "pull_request" ]]; then
+          echo "::warning::The GITHUB_TOKEN is required. Set the 'github_token' input parameter in order to publish summary."
+        elif [[ "${{ inputs.publish_summary }}" == "true" && "${{ github.event_name }}" == "pull_request" ]]; then
+          docker run --rm  \
+            -v ${{ github.action_path }}/templates:/templates \
+            -v ${{ inputs.workspace }}/${{ inputs.report_directory }}:/reports hairyhenderson/gomplate -d data=/reports/report.json -d results_master=/reports/report_master.json \
+            -f /templates/comment_summary.tmpl > comment_summary.md
+
+          OUTPUT=$(cat comment_summary.md)
+
+          if [[ "${{ inputs.cross_validation }}" == "true" ]]; then
+            pip3 install pytablewriter
+            OUTPUT="${OUTPUT}\n$(python3 ${{ github.action_path }}/scripts/cross_validation_results.py)"
+          fi
+
+          OUTPUT="${OUTPUT//$'\n'/'\n'}"
+
+          curl -X POST -s -H "Authorization: token ${{ inputs.github_token }}" -H "Accept: application/vnd.github.v3+json"  \
+          https://api.github.com/repos/${{ github.repository }}/issues/${{ github.event.number }}/comments -d "{\"body\":\"Commit: ${{ github.sha }}\n$OUTPUT\"}"
+
+        elif [[ "${{ github.event_name }}" != "pull_request" ]]; then
+          echo
+          echo "::warning::Publish Summary as a PR comment is available only for a pull request."
+          echo
+        else
+          echo
+          echo "Publish Summary as a PR comment is disabled. To enabled it set the 'publish_summary' input parameter to 'true'."
+          echo
+        fi
+
+    - name: Clean
+      shell: bash
+      run: |
+        rm -rf ${{ inputs.workspace }}/${{ inputs.report_directory }}/report_master.json
diff --git a/examples/report/report.csv b/examples/report/report.csv
new file mode 100644
index 0000000..1b444f0
--- /dev/null
+++ b/examples/report/report.csv
@@ -0,0 +1,5 @@
+name,macro_avg__f1-score,macro_avg__precision,macro_avg__recall,macro_avg__support,micro_avg__f1-score,micro_avg__precision,micro_avg__recall,micro_avg__support,weighted_avg__f1-score,weighted_avg__precision,weighted_avg__recall,weighted_avg__support,accuracy
+default/Rasa Demo - Configuration 1/entity_prediction,0.8439612155511906,0.8928145686878677,0.8505867162712345,1451,0.8802499132245749,0.8867132867132868,0.8738800827015851,1451,0.8795120459917696,0.8962881808799879,0.8738800827015851,1451,n/a
+default/Rasa Demo - Configuration 1/intent_classification,0.39396343663157196,0.6152770784589695,0.3412664077499422,4950,0.8144018583042975,0.9579234972677596,0.7082828282828283,4950,0.7571805042281962,0.89504509788388,0.7082828282828283,4950,n/a
+default/Rasa Demo - Configuration 1/response_selection,0.6128055852643406,0.6404843224661372,0.6527552082927716,1981,n/a,n/a,n/a,n/a,0.77835090522873,0.7812937731413816,0.8278647147905098,1981,0.8278647147905098
+default/Rasa Demo - Configuration 1/story_prediction,0.820486111111111,0.822530864197531,0.8363095238095238,189,n/a,n/a,n/a,n/a,0.915299823633157,0.9200470311581422,0.9206349206349206,189,0.9206349206349206
diff --git a/examples/report/report.json b/examples/report/report.json
new file mode 100644
index 0000000..56a9f2f
--- /dev/null
+++ b/examples/report/report.json
@@ -0,0 +1,76 @@
+{
+  "default": {
+    "Rasa Demo - Configuration 1": {
+      "entity_prediction": {
+        "macro_avg": {
+          "f1-score": 0.8439612155511906,
+          "precision": 0.8928145686878677,
+          "recall": 0.8505867162712345,
+          "support": 1451
+        },
+        "micro_avg": {
+          "f1-score": 0.8802499132245749,
+          "precision": 0.8867132867132868,
+          "recall": 0.8738800827015851,
+          "support": 1451
+        },
+        "weighted_avg": {
+          "f1-score": 0.8795120459917696,
+          "precision": 0.8962881808799879,
+          "recall": 0.8738800827015851,
+          "support": 1451
+        }
+      },
+      "intent_classification": {
+        "macro_avg": {
+          "f1-score": 0.39396343663157196,
+          "precision": 0.6152770784589695,
+          "recall": 0.3412664077499422,
+          "support": 4950
+        },
+        "micro_avg": {
+          "f1-score": 0.8144018583042975,
+          "precision": 0.9579234972677596,
+          "recall": 0.7082828282828283,
+          "support": 4950
+        },
+        "weighted_avg": {
+          "f1-score": 0.7571805042281962,
+          "precision": 0.89504509788388,
+          "recall": 0.7082828282828283,
+          "support": 4950
+        }
+      },
+      "response_selection": {
+        "accuracy": 0.8278647147905098,
+        "macro_avg": {
+          "f1-score": 0.6128055852643406,
+          "precision": 0.6404843224661372,
+          "recall": 0.6527552082927716,
+          "support": 1981
+        },
+        "weighted_avg": {
+          "f1-score": 0.77835090522873,
+          "precision": 0.7812937731413816,
+          "recall": 0.8278647147905098,
+          "support": 1981
+        }
+      },
+      "story_prediction": {
+        "accuracy": 0.9206349206349206,
+        "macro_avg": {
+          "f1-score": 0.820486111111111,
+          "precision": 0.822530864197531,
+          "recall": 0.8363095238095238,
+          "support": 189
+        },
+        "weighted_avg": {
+          "f1-score": 0.915299823633157,
+          "precision": 0.9200470311581422,
+          "recall": 0.9206349206349206,
+          "support": 189
+        }
+      }
+    }
+  }
+}
diff --git a/package.json.tmp b/package.json.tmp
new file mode 100644
index 0000000..ac7f9da
--- /dev/null
+++ b/package.json.tmp
@@ -0,0 +1,3 @@
+{
+  "version": "_version_"
+}
diff --git a/scripts/cross_validation_results.py b/scripts/cross_validation_results.py
new file mode 100644
index 0000000..fd95eaa
--- /dev/null
+++ b/scripts/cross_validation_results.py
@@ -0,0 +1,70 @@
+from pytablewriter import MarkdownTableWriter
+import json
+import os
+
+result_dir = os.environ["RESULT_DIR"]
+
+def intent_table():
+    writer = MarkdownTableWriter()
+    writer.table_name = "Intent Cross-Validation Results"
+
+    with open(f"{result_dir}/intent_report.json", "r") as f:
+        data = json.loads(f.read())
+
+    cols = ["support", "f1-score", "confused_with"]
+    writer.headers = ["class"] + cols
+
+    data.pop("accuracy", None)
+    classes = list(data.keys())
+   
+    classes.sort(key=lambda x: data[x].get("support", 0), reverse=True)
+
+    def format_cell(data, c, k):
+        if not data[c].get(k):
+            return "N/A"
+        if k == "confused_with":
+            return ", ".join([f"{k}({v})" for k, v in data[c][k].items()])
+        else:
+            return data[c][k]
+
+    writer.value_matrix = [
+        [c] + [format_cell(data, c, k) for k in cols] for c in classes
+    ]
+
+    return writer.dumps()
+
+
+def entity_table():
+
+    writer = MarkdownTableWriter()
+    writer.table_name = "Entity Cross-Validation Results"
+
+    with open(f"{result_dir}/DIETClassifier_report.json", "r") as f:
+        data = json.loads(f.read())
+
+    cols = ["support", "f1-score", "precision", "recall"]
+    writer.headers = ["entity"] + cols
+
+    classes = list(data.keys())
+    classes.sort(key=lambda x: data[x]["support"], reverse=True)
+
+    def format_cell(data, c, k):
+        if not data[c].get(k):
+            return "N/A"
+        else:
+            return data[c][k]
+
+    writer.value_matrix = [
+        [c] + [format_cell(data, c, k) for k in cols] for c in classes
+    ]
+
+    return writer.dumps()
+
+
+intents = intent_table()
+entities = entity_table()
+
+
+print(intents)
+print("\n\n\n")
+print(entities)
diff --git a/scripts/generate_results_report.py b/scripts/generate_results_report.py
new file mode 100644
index 0000000..79d3bc7
--- /dev/null
+++ b/scripts/generate_results_report.py
@@ -0,0 +1,62 @@
+# Collect the results of the various model test runs
+import json
+import os
+
+SUMMARY_FILE = os.environ["SUMMARY_FILE"]
+CONFIG = os.environ["CONFIG"]
+DATASET = os.environ["DATASET_NAME"]
+task_mapping = {
+    "story_report.json": "story_prediction",
+    "intent_report.json": "intent_classification",
+    "CRFEntityExtractor_report.json": "entity_prediction",
+    "DIETClassifier_report.json": "entity_prediction",
+    "response_selection_report.json": "response_selection",
+}
+
+
+def generate_json(file, task, data):
+    if not DATASET in data:
+        data = {DATASET: {CONFIG: {}}, **data}
+    elif not CONFIG in data[DATASET]:
+        data[DATASET] = {CONFIG: {}, **data[DATASET]}
+
+    data[DATASET][CONFIG] = {
+        **data[DATASET][CONFIG],
+    }
+
+    data[DATASET][CONFIG][task] = {**read_results(file)}
+
+    return data
+
+
+def read_results(file):
+    with open(file) as json_file:
+        data = json.load(json_file)
+
+        keys = ["accuracy", "weighted avg", "macro avg", "micro avg"]
+        key_mapping = {
+            "weighted avg": "weighted_avg",
+            "macro avg": "macro_avg",
+            "micro avg": "micro_avg",
+            "accuracy": "accuracy"
+        }
+        result = {key_mapping[key]: data[key] for key in keys if key in data}
+
+    return result
+
+
+if __name__ == "__main__":
+    data = {}
+    if os.path.exists(SUMMARY_FILE):
+        with open(SUMMARY_FILE) as json_file:
+            data = json.load(json_file)
+
+    for dirpath, dirnames, files in os.walk(os.environ["RESULT_DIR"]):
+        for f in files:
+            if f not in task_mapping.keys():
+                continue
+
+            data = generate_json(os.path.join(dirpath, f), task_mapping[f], data)
+
+    with open(SUMMARY_FILE, "w") as f:
+        json.dump(data, f, sort_keys=True, indent=2)
diff --git a/scripts/json_to_csv.py b/scripts/json_to_csv.py
new file mode 100644
index 0000000..e0fdedd
--- /dev/null
+++ b/scripts/json_to_csv.py
@@ -0,0 +1,55 @@
+import csv, json, sys
+from pathlib import Path
+
+def flattenjson(b, delim):
+    val = {}
+    for i in b.keys():
+        if isinstance( b[i], dict ):
+            get = flattenjson( b[i], delim )
+            for j in get.keys():
+                val[ i + delim + j ] = get[j]
+        else:
+            val[i] = b[i]
+
+    return val
+
+
+header = [
+    "macro_avg",
+    "micro_avg",
+    "weighted_avg",
+]
+
+if sys.argv[1] is not None and sys.argv[2] is not None:
+    fileInput = sys.argv[1]
+    fileOutput = sys.argv[2]
+    inputFile = Path(fileInput)
+
+    dataJSON = json.loads(inputFile.read_text())
+    data = flattenjson(dataJSON, "__")
+
+    with open(fileOutput, "w") as file:
+        csv_file = csv.writer(file)
+
+        header_names = []
+        for header_name in header:
+            for field in ["f1-score", "precision", "recall", "support"]:
+                header_names.append(header_name+"__"+field)
+        header_names.append("accuracy")
+
+        csv_file.writerow(["name"] + header_names)
+
+        for data_name in dataJSON:
+            for configuration in dataJSON[data_name]:
+                for row in dataJSON[data_name][configuration]:
+                    data = dataJSON[data_name][configuration]
+                    row_data = flattenjson(data[row], "__")
+
+                    result = [f"{data_name}/{configuration}/{row}"]
+                    for field in header_names:
+                        try:
+                            result.append(row_data[field])
+                        except KeyError:
+                            result.append("n/a")
+
+                    csv_file.writerow(result)
diff --git a/templates/comment_summary.tmpl b/templates/comment_summary.tmpl
new file mode 100644
index 0000000..2468042
--- /dev/null
+++ b/templates/comment_summary.tmpl
@@ -0,0 +1,91 @@
+{{- /*
+
+The template reads a file with a report (the report file is available
+as an artifact in the model regression tests workflow) and returns
+a markdown table with a summary of the tests.
+
+*/ -}}
+{{- /*
+
+The accuracy_tpl template returns data for the accuracy field.
+
+*/ -}}
+{{ define "accuracy_tpl" -}}
+{{- if has .master "accuracy" -}}
+{{ printf "%.4f" .branch.accuracy }} ({{ printf "%.2f" (.master.accuracy | math.Sub .branch.accuracy) }})
+{{- else -}}
+{{ printf "%.4f" .branch.accuracy }} (`no data`)
+{{- end -}}
+{{- end -}}
+{{- /*
+
+The micro_avg_tpl template returns data for the "micro avg" field.
+
+*/ -}}
+{{ define "micro_avg_tpl" -}}
+{{- if has .master.micro_avg "f1-score" -}}
+{{ printf "%.4f" (index .branch.micro_avg "f1-score") }} ({{ printf "%.2f" ((index .master.micro_avg "f1-score") | math.Sub (index .branch.micro_avg "f1-score")) }})
+{{- else -}}
+{{ printf "%.4f" (index .branch.micro_avg "f1-score") }} (`no data`)
+{{- end -}}
+{{- end -}}
+{{- /*
+
+Render Markdown with results.
+
+*/ -}}
+{{- $results_master := (datasource "results_master") -}}
+{{ range $dataset, $config := (datasource "data")}}
+{{- $dataset_master := (index $results_master $dataset) -}}
+Data: `{{$dataset}}`
+
+| Configuration | Intent Classification Micro F1 | Entity Recognition Micro F1 | Response Selection Micro F1 | Story Recognition Micro F1 |
+|---------------|-----------------|-----------------|-------------------|-------------------|
+{{ range $config_name, $config_data := $config -}}
+| `{{ $config_name }}` |
+{{- if has $config_data "intent_classification" -}}
+{{- /* Use results from branch if there is lack of data doesn't exist in results from scheduled version */ -}}
+{{- $intent_class_master := $config_data.intent_classification -}}
+{{- if has $dataset_master $config_name -}}
+{{- $intent_class_master = (index $dataset_master $config_name).intent_classification -}}
+{{- end -}}
+{{- $intent_class := index $config_data.intent_classification -}}
+{{- if has (index $intent_class "micro_avg") "f1-score" -}}{{ template "micro_avg_tpl" (dict "branch" $intent_class "master" $intent_class_master) }}{{- else if has $intent_class "accuracy" -}}{{ template "accuracy_tpl" (dict "branch" $intent_class "master" $intent_class_master) }}{{- else -}}`no data`{{- end -}} |
+{{- else -}}
+`no data`|
+{{- end -}}
+{{- if has $config_data "entity_prediction" -}}
+{{- /* Use results from branch if there is lack of data doesn't exist in results from scheduled version */ -}}
+{{- $entity_class_master := index $config_data.entity_prediction -}}
+{{- if has $dataset_master $config_name -}}
+{{- $entity_class_master = (index $dataset_master $config_name).entity_prediction -}}
+{{- end -}}
+{{- $entity_class := $config_data.entity_prediction -}}
+{{- if has (index $entity_class "micro_avg") "f1-score" -}}{{ template "micro_avg_tpl" (dict "branch" $entity_class "master" $entity_class_master) }}{{- else if has $entity_class "accuracy" -}}{{ template "accuracy_tpl" (dict "branch" $entity_class "master" $entity_class_master) }}{{- else -}}`no data`{{- end -}} |
+{{- else -}}
+`no data`|
+{{- end -}}
+{{- if has $config_data "response_selection" -}}
+{{- /* Use results from branch if there is lack of data doesn't exist in results from scheduled version */ -}}
+{{- $response_class_master := index $config_data.response_selection -}}
+{{- if has $dataset_master $config_name -}}
+{{- $response_class_master = (index $dataset_master $config_name).response_selection -}}
+{{- end -}}
+{{- $response_class := $config_data.response_selection -}}
+{{- if has (index $response_class "micro_avg") "f1-score" -}}{{ template "micro_avg_tpl" (dict "branch" $response_class "master" $response_class_master) }}{{- else if has $response_class "accuracy" -}}{{ template "accuracy_tpl" (dict "branch" $response_class "master" $response_class_master) }}{{- else -}}`no data`{{- end -}} |
+{{- else -}}
+`no data`|
+{{- end -}}
+{{- if has $config_data "story_prediction" -}}
+{{- /* Use results from branch if there is lack of data doesn't exist in results from scheduled version */ -}}
+{{- $story_class_master := index $config_data.story_prediction -}}
+{{- if has $dataset_master $config_name -}}
+{{- $story_class_master = (index $dataset_master $config_name).story_prediction -}}
+{{- end -}}
+{{- $story_class := $config_data.story_prediction -}}
+{{- if has (index $story_class "micro_avg") "f1-score" -}}{{ template "micro_avg_tpl" (dict "branch" $story_class "master" $story_class_master) }}{{- else if has $story_class "accuracy" -}}{{ template "accuracy_tpl" (dict "branch" $story_class "master" $story_class_master) }}{{- else -}}`no data`{{- end -}} |
+{{- else -}}
+`no data`|
+{{- end }}
+{{end}}
+{{end}}
diff --git a/templates/report_to_json.tmpl b/templates/report_to_json.tmpl
new file mode 100644
index 0000000..c63078d
--- /dev/null
+++ b/templates/report_to_json.tmpl
@@ -0,0 +1 @@
+{{- datasource "report" | toJSON -}}