diff --git a/tercen/util/io.py b/tercen/util/io.py
new file mode 100644
index 0000000..ada5b3a
--- /dev/null
+++ b/tercen/util/io.py
@@ -0,0 +1,368 @@
+from tercen.model.impl import \
+      InMemoryRelation, FileDocument, \
+        CompositeRelation, SimpleRelation,\
+        CSVTask, InitState
+
+import tercen.util.helper_functions as utl
+
+import os, tempfile, string, random, base64, hashlib, pickle, gzip, time, sys
+import pandas as pd
+
+from http.client import IncompleteRead
+
+
+def block_stdout():
+    sys.stdout = open(os.devnull, 'w')
+
+def enable_stdout():
+    sys.stdout = sys.__stdout__
+
+def filter_by_type(vec, type):
+    outVec = []
+    for o in vec:
+        if isinstance(o, type):
+            outVec.append(o)
+
+    return outVec
+
+def get_inmemory_relations(relation):
+    relations = []
+
+    if isinstance(relation, CompositeRelation):
+        rels = get_inmemory_relations(relation.mainRelation)
+        [relations.append(r) for r in rels]
+
+        for jo in relation.joinOperators:
+            rels = get_inmemory_relations(jo.rightRelation)
+            [relations.append(r) for r in rels]
+
+    elif isinstance(relation, InMemoryRelation):
+        relations.append(relation)
+    elif not isinstance(relation, SimpleRelation):
+        rels = get_inmemory_relations(relation.relation)
+        [relations.append(r) for r in rels]
+
+    return relations
+
+def get(vec, idxVec):
+    newVec = []
+    [newVec.append(vec[i]) for i in idxVec]
+    return newVec
+
+def where(vec):
+    return [i for i, x in enumerate(vec) if x]
+
+def get_document_id(queryRelation, aliasId, colName):
+    inMemRels = get_inmemory_relations(queryRelation)
+
+    for rel in inMemRels:
+        tbl = rel.inMemoryTable
+        
+        documentIds = get(tbl.columns, where([c.name == ".documentId" for c in tbl.columns ]))
+        # documentAliasIds = get(tbl.columns, where([c.name == "documentId" for c in tbl.columns ]))
+        documentAliasIds = get(tbl.columns, where([c.name == colName for c in tbl.columns ]))
+
+        if not documentIds is None and not documentAliasIds is None:
+            idx = where([id == aliasId for id in documentAliasIds[0].values ])
+            if not idx is None and len(idx) > 0:
+                return documentIds[0].values[idx[0]]
+
+    return None
+
+def get_gcs_filepaths(path):
+    paths = []
+    fs = os.listdir(path)
+
+    for f in os.listdir(path):
+        if os.path.isdir(path + "/" + f):
+            newPaths = get_gcs_filepaths(path + "/" + f)
+            [paths.append(p) for p in newPaths]
+        elif f.endswith(".gcs"):
+            paths.append(path + "/" + f)
+
+    return paths
+
+def read_in_chunks(file_object, chunk_size=1024 * 1024):
+    while True:
+        data = file_object.read(chunk_size)
+        if not data:
+            break
+        yield data
+
+def filter_by_base_folder(context, objList, baseFolder):
+    filteredList = []
+    for obj in objList:
+        objBaseFolder = get_folder_structure(obj.folderId, context).split("/")[0]
+        if objBaseFolder == baseFolder:
+            filteredList.append(obj)
+
+    return filteredList
+    
+def get_data(context, fileDoc):
+    maxTries = 10
+    downloadTry = 1
+    downloadSuccessful = False
+
+    data = None
+    while(downloadTry < maxTries):
+        try:
+            print("Downloading {} [Try {}]".format(fileDoc.name, downloadTry))
+            resp = context.context.client.fileService.download(fileDoc.id)
+
+
+            with gzip.open(resp, 'rb') as gFile:
+                data = gFile.read()
+                downloadSuccessful = True
+                break
+        except IncompleteRead:
+            print("Download failed. Trying again in 5 seconds.")
+            downloadTry += 1
+            time.sleep(5)
+
+
+    if not downloadSuccessful or data is None:
+        raise RuntimeError("Failed to download or extract {}".format(fileDoc.name))
+
+    return pickle.loads(data)
+
+def get_folder_structure(folderId, context):
+    folderName = ''
+    if folderId != '':
+        folder = context.context.client.folderService.get(folderId)
+        parentName = get_folder_structure(folder.folderId, context) 
+
+        if parentName != "/":
+            folderName += parentName + folder.name + "/"
+        else:
+            folderName += folder.name + "/"
+    
+    return folderName
+
+def get_filenames_with_folder(fileDocumentList, context ):
+    filenames = []
+    for fd in fileDocumentList:
+        filenames.append(get_folder_structure(fd.folderId, context) + fd.name)
+
+    return filenames
+
+
+
+def get_files_document_by_extension(fileDocumentList, ext):
+    fileDocuments = []
+
+    for fd in fileDocumentList:
+        if fd.name.endswith(ext):
+            fileDocuments.append(fd)
+
+    return fileDocuments
+
+
+def get_file_document_by_name(fileDocumentList, name):
+    fileDocument = None
+
+    for fd in fileDocumentList:
+        if fd.name == name:
+            fileDocument = fd
+            break
+
+    return fileDocument
+
+
+def find_in_dataframe(cols, value):
+        
+    colNames = cols.columns
+    for col in colNames:
+        if value in cols.select(col).to_numpy():
+            return True
+        
+    return False
+
+
+def get_filedoc_by_fullname(fileDocs, fullNameList, context):
+    outFileDocs = []
+    
+    if not isinstance(fullNameList, list):
+        fullNameList = [fullNameList]
+
+    for fd in fileDocs:
+        fdName = get_folder_structure(fd.folderId, context) + fd.name
+        if fdName in fullNameList:
+            outFileDocs.append(fd)
+        
+
+    if len(outFileDocs) == 1:
+        outFileDocs = outFileDocs[0]
+
+    return outFileDocs
+
+
+def download_filedocs(fileDocs, context, ext="" ):
+    savedFilePaths = []
+    baseDir = tempfile.gettempdir() + \
+              '/' +\
+              ''.join(random.choices(string.ascii_uppercase + string.digits, k=4))
+    
+    if not os.path.exists(baseDir):
+        os.mkdir(baseDir)
+
+    if not isinstance(fileDocs, list):
+        fileDocs = [fileDocs]
+    for fd in fileDocs:
+        fname = baseDir + '/' + \
+                ''.join(random.choices(string.ascii_uppercase + string.digits, k=8)) + \
+                ext
+        resp = context.context.client.fileService.download(fd.id)
+        #touch
+        f = open(fname, "wb")
+        f.close()
+
+        with open(fname, "ab") as file:
+            for chunk in read_in_chunks(resp):
+                file.write(chunk)
+
+        savedFilePaths.append(fname)
+
+    return savedFilePaths
+
+def text_to_markdown_df(filename, txt):
+
+    mimetype = "text/markdown"
+    txt = txt.replace("[", "**")\
+                .replace("]", "**")\
+                .replace("\n", "\n\n")
+
+    txtBytes = txt.encode("utf-8")
+
+    output_str = [base64.b64encode(txtBytes)]
+
+    outs = output_str[0].decode('utf-8')
+    txtDf = pd.DataFrame({
+        "filename":[filename],
+        "mimetype":[mimetype],
+        ".content":[outs]
+    })
+
+    return txtDf
+
+
+def export_to_project_as_csv(context, df, fname, projectId, folderId, user, timestamp=None ):
+    fname = "{}.csv".format(fname)
+
+    if not timestamp is None:
+        folderPath = get_folder_structure(folderId, context) + "/" + timestamp
+        newFolder = context.context.client.folderService.getOrCreate(projectId, folderPath)
+        folderId = newFolder.id
+
+    context.log("Exporting {}: Writing temp file".format(fname))
+
+
+    file = FileDocument()
+    file.name = fname.split("/")[-1]
+    file.acl.owner = user 
+    file.projectId = projectId
+    
+    file.folderId = folderId
+    file.metadata.contentEncoding = "application/octet-stream"
+
+    context.log("Exporting {}: Uploading".format(fname))
+    fileDoc = context.context.client.fileService.uploadTable(file, \
+                                                              utl.dataframe_to_table(df, values_as_list=True)[0].toJson())
+    
+    task = CSVTask()
+    task.state = InitState()
+    
+    task.fileDocumentId =  fileDoc.id
+    task.projectId = projectId
+    task.owner = user
+
+    task = context.context.client.taskService.create( task )
+    context.context.client.taskService.runTask(task.id)
+    taskDone = context.context.client.taskService.waitDone(task.id)
+
+    # Move to correct folder
+    sch = context.context.client.tableSchemaService.get(taskDone.schemaId)
+    sch.folderId = folderId
+    context.context.client.tableSchemaService.update(sch)
+
+    return df
+
+def export_to_project(context, data, fname, projectId, folderId, user, timestamp=None, compression=1, fileExt="gz"):
+    fname = "{}.{}".format(fname, fileExt)
+
+    if not timestamp is None:
+        folderPath = get_folder_structure(folderId, context) + "/" + timestamp
+        newFolder = context.context.client.folderService.getOrCreate(projectId, folderPath)
+        folderId = newFolder.id
+
+    context.log("Exporting {}: Writing temp file".format(fname))
+    with gzip.open(fname, 'wb', compresslevel=compression) as f:
+        pickle.dump(data, f)
+    
+    del data
+    data = []
+
+    file = FileDocument()
+    file.name = fname.split("/")[-1]
+    file.acl.owner = user 
+    file.projectId = projectId
+    
+    file.folderId = folderId
+    file.metadata.contentEncoding = "gzip"
+
+    context.log("Exporting {}: Uploading".format(fname))
+    context.context.client.fileService.uploadFromFile(file, fname)
+    
+
+    return data
+
+
+def append_img_to_df(df, imagePath, pltCi=0):
+    if df is None:
+        df = utl.image_file_to_df(imagePath)
+        df.insert(0, ".ri", int(pltCi))
+    else:
+        tmpDf = utl.image_file_to_df(imagePath)
+        tmpDf.insert(0, ".ri", int(pltCi))
+        df = pd.concat([df, tmpDf])
+    return df
+
+
+def pickle_file_to_df(file_path):
+    filename = os.path.basename(file_path)
+    mimetype = "application/octet-stream"
+
+    output_str = []
+
+    with open(file_path, mode="rb") as f:
+        fc = f.read()
+        output_str.append([base64.b64encode(fc)])
+
+
+    o = output_str[0][0]
+
+    outs = o.decode('utf-8')
+    pklDf = pd.DataFrame({
+        "filename":[filename],
+        "mimetype":[mimetype],
+        ".content":[outs]
+    })
+
+    return pklDf
+
+
+def string_to_df(txt, filename):
+    import hashlib, base64
+    mimetype = "text/markdown"
+
+    txtBytes = txt.encode("utf-8")
+
+    output_str = [ base64.b64encode(txtBytes) ]
+    outs = output_str[0].decode('utf-8')
+    txtDf = pd.DataFrame({
+        "filename":[filename],
+        "mimetype":[mimetype],
+        ".content":[outs],
+    })
+
+    return txtDf
+